[sdiy] Various colors of noise.

[Donald Tillman]

On Mar 3, 2026, at 1:58 PM, Mike Beauchamp <list at mikebeauchamp.com> wrote:
> 
> 
> Maybe take all of the different noises, and feed them into an interpolating scanner circuit (like this one: https://till.com/articles/scanner/ ) and from there you could control the output noise colour (and anything inbetween) with the wandering CV.

That interpolating scanner article is crazy!

> I wonder if the different noise colours all need to be derived from a single source for the crossfading to work.

Theoretically the opposite; a single noise source allows the possibility of phase cancellation between two different filter functions, while separate random sources guarantee no phase alignment.  That said, it's unlikely to be an issue.

Noise isn't my speciality, I'm usually trying to get rid of it, but I believe that...

A simple noise source, such as just generating a series of random values, should be "white noise", with a constant power level over any given bandwidth in Hz, regardless of frequency.  So the power content from 0 to 1KHz should be the same as 1KHz to 2Khz, and 2KHz to 3KHz, and so forth.

Our sense of hearing works over logarithmic frequency, per octave, so we expect a constant power level per octave.  That's called "pink noise".  

Starting with white noise, we need a low-pass filter that drops 3dB for each octave to make pink noise.  A half-integrator, if you will.  You can approximate it with a few low-pass shelf sections spaced out.

And the 3dB/Oct integrator was adopted as the way to go from one color of noise to another.  White to pink, pink to red, and so forth.  Or the other direction; white to blue.

So a noise generator with a continuous color control would basically involve a variable slope filter.  (Which is a pretty cool thing on its own, anyway; though more for mixing and regular instruments than synthesis.)

  -- Don
--
Donald Tillman, Palo Alto, California
https://till.com