To avoid mix ups it is important to stress that the kind of compression that we will be analyzing in this section is totally different to the one acting upon the signal's amplitude, which is another type of compression described in a separate section [Compressor ] .
Of all the manipulations that can be carried out on a sampled audio signal, data compression deserves particular attention and we will now take a look at its specific characteristics. We have seen how the sampling process carried out following the CD standards (16 bit, 44.1 KHz) produces 172.26 KB per second, therefore even just a few seconds of music is the equivalent of a large amount of data.
In some contexts however, this can turn out to be a problem, as is the case when one wishes to store lots of tracks, or when one accesses a track that has been stored on the computer by using a remote workstation[32 ]. When working with many tracks, the greater the quantity of data we have to store, the larger the size of the support we need, and thus, the more expensive it will be. The transfer of data through a digital network takes place at a relatively low speed and therefore to transfer a single track-file takes a very long time and puts the network's resources under great strain. This is why we'll often end up compressing the audio signal data.
Before we step into compression and its details we need to remember that there are two different kinds of compression: loseless (without any loss of information) and lossy (with loss of information). The first kind of compression allows us to reconstruct the original data precisely; in other words, by applying a decompression to the compressed data, we recover exactly the original data we started off with. The second type of compression is rather more drastic, eliminating part of the data (the part that is considered of lesser importance) which means that after the decompression phase we will get an approximated version of the original data. Clearly, lossy compression gives considerably greater compression-ratio values[33 ] compared to loseless compression. However, there are cases in which lossy compression will not do as for example when decompressing a text document where we need to retrieve the exact original data. When it comes to images[34 ] or sounds however, lossy compression does the job, because approximated values after the decompression phase are pretty similar (though not identical) to the original.
[32 ] A remote workstation consists in a computer connected to an I.T. network (which can be composed even just by one computer).
[33 ] This is a measuring unit that identifies the amount of compression that has been reached. It is obtained by calculating the ratio between the amount of original data and the amount of compressed data. For example if a music track occupies of 3.45 MB before compression and after compression is reduced to 890 KB the ratio is 3.969: in this case we've carried out a compression with a compression ratio of 4:1 (from 4 to 1).
[34 ] The compression formats for images are numerous and have lots of different scopes : GIF (used for images generated by computers), JPEG (used for images that are imported from the real world), TIFF (used for high-definition images) and others.