[sdiy] Sigma Delta DSP

[ASSI]

On Mittwoch, 1. März 2006 17:45, Dave Krooshof - dendriet.nl wrote:
> The downside however, is that I would not even know how to
> do gain properly in a bitstream, let alone filtering.

There's a number of PhD thesis' in that question... :-)

> Can anyone point us to information about bistream DSP?

You may actually have more luck looking up publications on how 
binary/trinary pulsed neural networks process information.

Paramount to DSP directly with pulse trains is to chose an efficient 
(and likely redundant) number system for each operation at hand (for 
integration and thresholding). That implies frequent number system and 
sample rate conversions. Multiplication and addition may both be done 
by XOR gates in the right representation, while other operation may be 
more efficiently implemented by splitting up the signal into several 
bitstreams that are processed in parallel before funneled together 
again. The second key ingredient IMHO is a number of mutually 
independent (white) noise sources with adjustable probability of the 
pulse density (the tricky part here is the seemingly innocent 
"independent" as you'd need lots of them). And now the best or worst 
part, depending on how you look at things: you gotta do this in an FPGA 
at least (ASIC would be even better), because the most common 
operations are wildly inefficient to do in software.


Achim.
-- 
+<[Q+ Matrix-12 WAVE#46 Neuron microQkb Andromeda XTk sonic heaven]>+

SD adaptations for Waldorf Q V3.00R3 and Q+ V3.54R2:
http://Stromeko.Synth.net/Downloads.html#WaldorfSDada