Sep 282004

[suffusion-widgets id='2']

A digital sound is actually quite big: Each second of a stereo CD contains 1,411,200 single informations (bits), or 176,400 Bytes, or 172.25 KBytes. On a computer this type of sound usually comes as a file with the extension “wav” on the PC and “aiff” on the Mac. Downloading a second of CD quality sound would last about 2 minutes on an average Internet connection with a 28 k modem. (A 14 k modem takes about twice as long…)

The same file type can also store sound information in lower quality. Three things can get worse: Stereo can become Mono (1 channel instead of 2), the “resolution” can be reduced from 16 bit to 8 bit or even 4 bit, and the “sample rate” can be reduced from 44.1 kHz to 22 kHz, 11 kHz or 5 kHz.

Any of these reductions are “divide by 2″ operations – and that’s what will happen to the file size! One second of sound (wav or aiff) in mono, 22 kHz, 8 bit is about 21.5 Kbytes. One second in mono, 5 kHz, 4 bit is about 2.7 Kbytes.

Doesn’t sound big? True – but doesn’t sound good either! And now consider this: 2.7 Kbytes is pretty much a dream performance of a 28 k modem connection to the Internet! It happens, but not very often. Realistically speaking, a “good” connection with a 28 k modem makes something between 1.5 and 2.3 Kbytes per second – not even enough for the bad sound at 5 kHz 4 bit.

While audio files can only become smaller by “compressing” them (trying to put more information in the same space) there is one sound format that’s naturally slim and trim: MIDI. The size of a MIDI file depends somewhat on how much action is in the music. Starting at about 100 Bytes per second, you will hardly find a MIDI file that’s more than 500 Bytes per second! Yep, that’s small.

So, why not stuff all the audio files into MIDI format and use this for the Internet? There’s a problem: MIDI can’t be recorded like audio, it must be “programmed”, and it’s limited to sounds that a “sound module” or “sound card” can generate. That means no human voices, no natural instruments – nothing that can only be recorded with a microphone or something like this.