MP3 compressed sound files are finally becoming popular with online music stores, now that DRM is on the way out. Although MP3 compression is not quite as efficient as AAC and WMA it is good enough. The files are much smaller than WAV files and audio CD data files – around one tenth of the size, depending on data rate. A song in the form of a 4Mb MP3 file would take up 40Mb if it was stored an an uncompressed a WAV file. That means the files are good for sending via broadband and they’re also ideal for systems with limited memory like personal MP3 players.
Digital audio works by passing the audio signal into an analogue to digital converter which takes thousands of measurements of the voltage level every second. The sound is sampled 44,100 times per second for CD audio. Each sample uses 16 bits for each channel so the sound can be accurately reproduced. Which adds up to an awful lot of bits per second.
To reduce the number of bits a second the data has to be compressed. There are lots of different ways of compressing signals. Compression systems fall into two groups – lossy and lossless. Lossless systems allow the original data to be accurately reconstructed. Lossy systems allow higher compression levels but don’t allow the original signal to be faithfully reproduced.
Now what’s all that got to do with MP3? Well MP3 files needed to be small but also to sound reasonably good which sounds pretty impossible. MP3’s design takes advantage of a feature of human psychoacoustics – loud sounds hide quiet ones. In short you don’t have to send the data for any quiet sounds that occur at a similar frequency to loud ones.
MP3 divides the audio frequency range up into lots of different bands for the analysis and compression. The sampling rate can also be reduced to further reduce the bit rate. Both technniques result in some loss of audio quality though.
The most common data rate for MP3 files is 128 kilobits per second, which is roughly equivalent to FM radio quality. Higher bit rates of 190 kilobits and above allow full hi-fi quality.
Currently most MP3 systems use fixed bit rate (FBR) sampling, which means an awful lot of bits wasted during silence and simpler sound sequences. It also means worse quality during complicated sequences. Variable bit rate (VBR) produces better quality sound for a given bit rate but uses more power to decode – a crucial difference for battery-powered devices.