Fraunhofer says that 128kbps is supposed to give "CD Quality sound"... So first, let's have a look at a real CD Quality file. Here is the frequency analysis of a 2 minutes long sound, part of "The Empire Strikes Back" (an orchestral movie theme) by John Williams.

As you can see, it has a 22kHz bandwidth because it's CD Quality audio, but you have less important information after 16kHz. Most of us can't hear a 16kHz sound, however, the whole perception of audio is often not perceptible by the human ear, but still by the brain. Even though we do not hear them, we sense more than we hear. Even when it's humanly impossible to hear a sound, it is still 'pleasing' to 'hear' them.  (This has a lot to do with listener fatigue, etc.)

If you don't agree with this, or never realized how high a 17Khz sound is, just try to create a 17kHz sine wave (use "generate tone" with Cool Edit for example) and listen to it. You'll hear nothing but your hi-fi set's noise.

------------------------------------------------------------------------

Much more interesting, here is the frequency analysis of an MP3 file, to be accurate a 128kbps MPEG1 Layer 3 file. It's the conversion of the original WAV file, compressed 12 times.

Below 16kHz, we have an exact match, the two signals are 100% identical (that's why the MP3 sounds great), but at 16kHz : everything stops. The MP3 compression seems to limit the bandwidth of this sound ! With the 128kbps bitrate (the most popular on the Internet, supposed to give "CD quality sound") we have absolutely nothing left above 16kHz !

This does not mean that "128kbps compression has always a 16kHz bandwidth". It's a lot more complex than that. There are two main conditions, that will be discussed in the "How MPEG Audio works" section of this site.

1.First of all, the signal components in the higher bands have to be strong enough so that they won't be covered by the signal components in the lower frequency range. If the model inside MP3 says that the highband components are covered by lower frequencies (this means that you'll never hear them), then the highband components will be discarded.

2.The MP3 process analyses the higher bands components, but even if one can hear them, MP3 takes only care of them if it has enough bandwidth left in the resulting bitstream (e.g. 128kBit/s). Then it weights all the components in the original signal and takes care of them as much as possible until the 128kBits are fully used.

------------------------------------------------------------------------

In audio tests that were conducted by the standards commission, the testers couldn't distinguish between the real DAT level samples and MP3 encoded samples at 256kbps. With Compression rate 6:1 (16 bits stereo sampled at 48 kHz is reduced to 256 kbits/sec) and optimal listening conditions, expert listeners could not distinguish between coded and original audio clips. (Subjective testing has been used extensively by MPEG and other organisation in order to grade codec quality.)

However, the bad news is that bigger bitrate... A reduction ratio of 6:1 is too low, and a 256kbps file is always two times bigger than a 128kbps one.

If a link on this page doesn't work, let me know It doesn't fuckin'work!

All this web page is designed by Dave Guile © 1998