Lena:

Gosh...happy Valentine's day to you too!

You might also be interested in these two files too. They show the effects of the use of "dither" in digital audio. Here's the complete post from another forum:

There is one thing that I have not seen mentioned in this discussion: that is the addition of dither to the LSB. This allows you to hear below the theoritical noise floor for a particular bit depth. I actually created a CD containing a series of tests that proves that this works, and does indeed allow the audio to be heard below the noise floor. I have included a couple samples from this CD here. If the dither is of the "noise shaped" variety, you can add an effective 2 bits to the nominal resolution of the digital word. Once you get near and beyond the 16 bit level, the "self noise" of the electrical system, microphone, and acoustic room noise create a natural dither.

The following samples use an 8 bit quantitazation level so that the effects of the quantitization process and of the dither can be clearly heard. For higher bit depths, the effects would be the same, but farther down in level. It consists of a short "excitation" of a 24 bit digital reverb, which was quantitized to the 8 bit level, with and without noise shaped dither. All processing was done in the digital domain.

Sample with noise shaped dither

Sample with no dither

Notice in the dithered example how the reverb tail can be clearly heard to fade BELOW the "hiss", which is the dither noise at the LSB level. The example without dither cuts off abruptly as soon as the level of the reverb tail reaches the limits of the noise floor for 8 bits, which is 48db. This abrupt cut-off was the reason some early CDs sounded like the low level details "fell off into a black hole" - dither was not in common use when digital audio CDs were first introduced. It is now routine practice to use dither in all recording and processing of digital audio. Some companies like Sony (with their "SuperBit Mapping" CDs) use noise shaped dither to increase the audible signal to noise range, allowing a 16 bit CD to have the effective resolution of an 18 bit recording.

[This message has been edited by soundhound (edited February 14, 2003).]