[sdiy] Uniformly distributed noise generator?

rburnett at richieburnett.co.uk rburnett at richieburnett.co.uk
Mon Jun 10 14:06:54 CEST 2013


> ...But I wonder if there isn't a specific circuit that inherently 
> creates
> uniformly distributed noise, similar to how we get white noise via the
> inherent properties of what ever semiconductor is used. I have a
> feeling something like this must exist, after all it's a basic concept
> in statistics, and a lot of stats got done on analog computers for
> many decades.

You want something analogue that inherently generates an even 
distribution of amplitudes between two rigid limits, and then NOTHING 
outside these limits?  That sort of distribution doesn't sound like an 
inherently natural behaviour to me.

BTW.  This paper from Jon Dattorro contains an interesting section on 
digital generation of noise:

https://ccrma.stanford.edu/~dattorro/EffectDesignPart3.pdf

It's a great paper in general, but section 8 is on noise generation in 
particular.  He discusses the audible repetitions (chuffing train) that 
occur with short-length LFSR implementations, and offers a selection of 
various length generator polynomials (LFSR tap choices.)  He also points 
out that a single bit from the LFSR gives spectrally white 1-bit noise, 
but explains why casting larger parts of the shift register into say a 
16-bit signed word can result in noise that is no longer spectrally 
flat.  The spectrum can then be corrected by a simple "whitening" 
filter.

-Richie,

PS. For those interested in DSP audio/music, all three of those 
Dattorro papers are worth grabbing and stashing for reference.


More information about the Synth-diy mailing list