[sdiy] Handy DIY tip #213/noise color

[Magnus Danielson]

From: Grant Richter <grichter at asapnet.net>
Subject: Re: [sdiy] Handy DIY tip #213/noise color
Date: Wed, 15 Jan 2003 01:25:01 -0600

> Signal to noise is kind of a misleading term.
> 
> In the case of a noise generator, the noise IS the signal, so any signal
> contained in the noise, would be considered "noise". A reversal of roles.

Exactly!

> For example, if you fail to bypass the supply to a reversed biased emitter
> noise generator, you can get other periodic signals from the power supply
> lines. These periodic signals are "noise" since they disturb the gaussian
> distribution of the noise (which is the signal of interest).

Actually, you allways have the overtone spectra from the supply, but the real
question is just how much damping there is.

A good way to measure the "quality" of noise should really be to measure in
dB the autocorrelation of the noise. Normal noise should not autocorrelate with
itself, but any non-noise signal in there would autocorrelate with itself, and
contribute to the autocorrelation. The lower autocorrelation number, the better
quality the noise have. Beyond a certain level I think you could consider it
for most applications to be _only_ noise.

However, I know there are issues with certain types of noise and statistics, so
ordinary autocorrelation might not have enought "strength". Noise is a mighty
strange thing you know.

Cheers,
Magnus