[sdiy] Walsh Generators

[Neil Johnson]

Hi Glen,

(about time I popped up!  been v.busy at work at the moment)

> Okay then, show me how you can input 16 higher-resolution values into
> Neil's Tcl Walsh Synthesizer, and end up with a saw wave having more 
> than
> 16 discrete steps.  :)

I think you are confusing sample rate with resolution.  You will see 
that the sliders (amplitude) can be moved in very small steps.  But 
there are only 16 samples.  So if you wanted to record this at 44.1kHz 
sample rate this would produce a sawtooth of 2.75kHz.

The only reason I made my little toy 16 steps long is because that's 
all that would fit on a webpage.  The Walsh synth I built for my 
Masters degree had 31 steps (ignoring the DC component).  I did once 
sketch out some plans for a 63-term Walsh synth, but so far haven't had 
the time to get it going.  Funnily enough, I get to play with FPGAs at 
work at the moment, so maybe one day...

I think the major point brought up so far is the issue of the user 
interface.  I think the most successful additive synth is the Hammond, 
with its drawbars mixing together a variety of fundamental waveforms.  
Getting the user interface right is the key.

Look at the Kurzweil K150FS additive synth (designed by Hal 
Chamberlain).
	http://www.vintagesynth.com/index.html

Another example being the Fairlight CMI, which also had a page for 
graphically drawing the spectral content of a waveform (i.e., defining 
the mix of sinewaves), which was then synthesized and loaded into the 
sampler.

Some more info here:
	http://www.soundonsound.com/sos/jun00/articles/synthsec.htm

Kawai dabbled a bit in additive synthesis with the K5, and latterly the 
K5000 (yes, I have a K5000, and it is rather good at Hammond sounds!).

Resolution and sample rate can be had for almost free nowadays with 
huge, cheap FPGAs.  I'm looking at using the smallest Spartan-3E on a 
project---it has 100,000 gate-equivalent, multipliers, block RAM, etc.

The challenge is how you control all this in a meaningful, useful and 
intuitive way.  Otherwise how can you make an instrument out of it?

On the issue of quality, I would do _all_ the math in the digital 
domain (you don't need multipliers, only add and subtract), then send 
the output waveform through a DAC for traditional analogue signal 
processing.

Neil
--
Neil Johnson <http://www.njohnson.co.uk>