Mad Man Rambles about Noise Generators

[Don Tillman]

   Date: Sun, 2 Jun 1996 11:00:09 -0700
   From: chordman at ix.netcom.com (Scott Gravenhorst, Synthaholic)
 
   I understand the math behind the number of possible states and the 
   doubling of them by adding another stage.  However, I was playing with 
   a system who's output is only one bit.  I was looking at the problem 
   from the standpoint of one half cycle at a time and given that each 
   half cycle lasts a certain amount of time, it could be said to have a 
   sort of instantaneous frequency characteristic that changes from half 
   cycle to half cycle, giving rise to 'random frequency'.  I looked at 
   the length of the shift register as determining the longest possible 
   half cycle.

Two words: FFT

(That's stolen from Metlay I think.  It's just such of great line!)

Doing an FFT analysis on various feedback shift register
configurations is a great way to deal with some of this.  It's easy
and quick.

Back to the topic.  I think you're referring to that large string of
0's (and later 1's) all in a row, right?  I'm pretty sure this is what
causes that thumpa effect.

If that's the case I guess there are two approaches; either avoid the
string of 0's, or use it to your advantage.

When I first proposed a VCO driven feedback shift register to AH a
long while ago, the idea was to make the harmonic structure of the VCO
waveform far more complex.  So for that, the string of 0's is just
more fundamental tone.

It sounds like you're using this more as a noise source, and you want
to get rid of the string of 0's.  Okay.  I'm guessing you tap off
just the last two SR stages.  If you tapped off other stages in the
middle, I beleve you'll eliminate that string of 0's.  Obvously you'll
want one tap at the very end.  Some combinations of taps drastically
reduce the length of the sequence, so you'll want to either do a
quickly simulation or bone up on error correction theory to be sure.

  -- Don