Re: KT Piano Demos on CD

[Craig Heath <craig@sco.COM>]

> From: Anders Hultman <anders.hultman@unisource.se>
> Date: 14 Dec 1995 10:31:18 GMT

> willker@chemie.uni-bremen.de (Wieland Willker) wrote:
> >Maybe one
> >can record the noise in the intervalls between the songs, calculate some
> >mean noise and subtract it from the songs? Any expert out there?

> That is not possible. Noise is, per definition, totally random. There is 
> no way to take away noise without taking away other things as well.

This is of course, technically true; however, it is possible to be a lot
more selective about filtering out noise frequencies than simply cutting
off the top end (i.e. turning the treble down).  The Windows shareware
utility COOL EDIT has a function which will do an FFT analysis of noise
in a content-free part of the signal (i.e. the gap between songs), and
will then filter the whole signal to remove all or part of the noise
frequencies.

I have experimented with this on "Disbelieving Angel", with a small
amount of success; the tape I have does not include much of the
in-between song gaps, however, and I'm sure a better job could be done.
There are two factors - getting a good sample of the signal-less noise,
and then the judgement of how much of the noise to remove traded off
against inevitable loss of parts of the true signal.

As an aside, you can use this technique to analyse say, the frequency
spectrum of the piano part, and then remove that from the track -
fun, but ultimately not very useful!  It is also _very_ slow on my
machine, taking over 1 hour real time for 1 minute of recorded signal.

		- Craig @ SCO near London.