[sdiy] Noise Distribution

[René Schmitz]

>What you say about shaping to other distributions is true, of course. But I 
>wonder if there isn't a further reason. If the distribution is not in some 
>sense equalized (white, pink, I don't care), then certain frequencies or 
>frequency bands will be emphasized. I think this would give an audible 
>characteristic coloration to the sound. It seems to me this would be 
>undesirable, just like coloration in a loudspeaker system.

I think thats true. As a thought experiment assume white gaussian noise 
feed it thru a waveshaper, lets say a comparator to get a binary amplitude 
distribution. One could argue that the individual "harmonics" or rather 
partials that make up the signal get distorted. I.e. more signal power at 
high frequencies. The noise sounds more "crispy". Similar with other 
distribution transformations. 

Yet, as one can shape a sound with more harmonics into one with less
harmonics 
(sineshaping from tri), I think that should be also possible with noise. 
My reasoning is that it is easier to shape from linear distribution to start, 
just as it is easier to shape from a triangle because your transfer curve is 
already the desired curve.

Cheers,
 René


-- 
uzs159 at uni-bonn.de
http://www.uni-bonn.de/~uzs159