[sdiy] Digital noise generation

[Magnus Danielson]

From: "Fredrik Carlqvist" <ifrc at iar.se>
Subject: RE: [sdiy] Digital noise generation
Date: Wed, 9 Mar 2005 08:26:39 +0100
Message-ID: <200503090726.IAA26565 at pjakkur.iar.se>

Hi Fredrik,

> All M-sequences produce every number between 1 and 2^N-1, but the tap
> positions and number will change the ordering. For some combinations, there
> will be several distinct pops or even brief tones somewhere in the cycle,
> others are smooth. If you listen to an unprocessed LFSR output you will soon
> be able to determine it's length by ear, at least if it's shorter than a
> minute.

Not all tap positions generate a Maximum Length Sequence, i.e. having 2^N-1
states. Actually, for some N there is even no sequence using only one "M"
tap, but you are forced to use several.

> It's not so much a matter of the number of taps, a 2-tap LFSR can be as
> smooth as a 12-tap. 
> 
> Using more than one bit from the LFSR output (while increasing the shifting
> speed accordingly) will not change the distribution, but probably remove som
> pops. 

Not if you XOR them, then you only phase-shift the sequence. If you just
add them you would would get a little dispersion of the pop-amplitude, but you
may just get a phase-shifted variant there too (recall, XOR is addition in
modulo-2).

Cheers,
Magnus