Mad Man Rambles about Noise Generators

[Scott Gravenhorst, Synthaholic]

Thinking out loud...

A while back I had posted a request for information about shiftregister
noise generators, specifically the tap points for feedback for sizes up
to 64 bits.  I got the required information thanks to this some kind
souls on this list.  At that time, there was a question (I can't
remember who) posed something like 'just how random do you need your
noise?'.  This was in response to a statement I made that I wanted to 
try a shift register perhaps as large as 64 bits.  I am currently in
the middle of the breadboarding of my voltage controlled noise 
generator which is going quite well, but the question kept gnawing at 
me.

It occurs to me that there is more to the size of the shift register
than simply the repeat time.  There is a question of spectrum size as
well.  It would seem to me that a shift register of 16 bits would be 
able to produce noise an octave lower than one of 8 bits, but that the 
high end of the spectrum would be controlled by the clock rate alone in 
both the 8 and 16 bit versions.  Assuming, of course, a constant and 
equal clock rate for both.  So, the highest (instantaneous) frequency 
produced would be when a 1010101... pattern is currently being shifted. 
 This would be one half the clock frequency.  The low end would be the 
clock frequency divided by 2 times the number of bits that comprise the 
shift register (well, not exactly, due to the exclusion of the one 
'illegal' state) because all of the bits could be low for N clocks and 
then high for N clocks giving a cycle time of 2N clocks.  (where N is 
the size of the shift register in flip flops).

So I would conclude that not only does one extend the repeat time by
increasing the number of flip flops, but also the bandwidth of the 
noise.

I am also going to experiment with a PLL attached to the noise 
generator output.  I will try both 1 and 2 poles (2 poles config with a 
capacitor range switch) and I will be interested in the demodulated 
output.  The voltage at the demod output should represent the 
instantaneous frequency, very much like an FM demodulator.  Since the 
raw digital noise is of constant amplitude, but changing frequency, and 
the demodulated output will be comprised of random amplitude as well as 
random frequency, I will expect that the 'color' and therefore the 
sound characteristics of the noise should be at least somewhat 
different, and should also sound different from digital noise passed 
through a lowpass filter.

I'll post what I find.



-- 
-- Scott G., Synthaholic

If at first you don't succeed, keep suckin' 'til you do suck seed.
                                                 -- Curly Howard
What else can happen?
       -- Howard (Never call me Howie) Bradley