[sdiy] VERY cool/bizarre: Buchla noise

[harrybissell]

Scott Gravenhorst wrote:  <snipp'd>

> I believe in LFSR noise generators, as long as they are at
> least 32 bits long.  If my math is right, that's around 4300
> seconds (71 minutes) when clocked at 1 MHz.  I've found that
> clocking them at far less than that still yields what sounds
> indistinguishable (to me) from real white noise.

The artifacts would probably not be distinguishable to the
naked ear :^P   If you had a very very narrow bandpass filter...
you could find missing frequencies.  But perceptually you will
hear a good number of simulataneous frequencies as "noise"

The fact that we use varying clock rates to make 'pitched' noise
shows that we can perceive the difference... but if you did not
have another reference (like a different noise source played just
prior to another - or swept...) you would just accept the result
as being 'noise'.


> Of the
> features of this method are no popcorn noise, no need for
> selection/rejection of components and a predictable, large
> output level without need for amplification.  And it's a
> cheap thing to build and has no critical layout needs.  The
> more bits the better.  I have a Xilinx PDF with a table
> detailing XOR taps for maximum length LFSRs up to 168 bits.

Maybe we should make a demo board that lets you step through
lobger and longer pseudo-random sequences.  Good job for our
FPGA-totin' friends.   I suspect that we would find the noise
unacceptable
to a certain point, and then notice no further changes after that

H^) harry