[sdiy] analog loses another niche

[Mike Pepper]

> One fine day I got clever and hooked up a LFSR to an RF spectrum analyzer.
I thought with averaging turned on I would see a straight line. Instead what
I saw was a picket fence of peaks, or in RF parlance, a comb generator. With
an 8-bit LFSR the peaks were quite pronounced, with a 16-bit they were so
close together it was tough to see them, but they were there. Bottom line:
digital noise really does sound different, unlike CDs that have been
purified with flashing LEDs.
>
Nice move. If I remember my LFSR theory, you get  bandwidth limited noise,
the bandwidth being roughly the clock signal divided by the sequence length,
(assuming a maximal length sequence). The comb effect should repeat at
multiples of the sequence length. To use digital noise in a an analog
environment, you need to put an anti-aliasing filter on the output.

The paper here has some stuff on the subject, (I can't find the hardcopy
article I used to use).

http://www.imeko.org/publications/iwadc-2007/IMEKO-IWADC-2007-F046.pdf

You can make a fun sort of digital oscillator by using a high speed clock
into an LFSR  having switch selectable taps, and sending the output bits via
switched equal resistors into a summing amp. Gives a wide range of funny
waves or LFO effects, but tuning is rather 'by ear', with some of the
sequences. Oh, and a reset button helps: switching feedback can have the
effect of setting a new cycle using the existing bits in the register, OK,
but it may suddenly stop or 'jam', (get stuck with  all 1's or all 0's), or
produce a sequence you can never replicate. I've been playing with an LFSR
add on to my Klee sequencer, to produce sorts of pseudo random arpeggios,
though sound samples will be a while yet, (I need to get the studio part of
my workroom up and running!).

||\/||ike