[sdiy] Additive VCO

[ASSI]

On Monday 08 March 2004 00:25, M.A. Koot wrote:
> My initial idea is by having one expo-converter, having the output
> devided over several Opamp-adders (to create the harmonic-offset).

The expo output is normally a current, so it'd probably work better if 
you mirror it to each VCO and taper the mirror ratio according to the 
harmonic series. Alternatively you can take an expo for each VCO 
(they'd need to track well) and just add a constant voltage offset at 
the input. This gives the additional benefit that you can modulate the 
ratio by a CV.

> The things I'm troubling with though, is how many harmonics (i.e.
> Sine-VCO's) should I have at least to create a decent Additive VCO? I
> was thinking of about 16 Sine-VCO's.

I've gotten decent results out of just five partials, however I would 
not use pure sines only. If you can provide for rectangle (perhaps with 
PWM) and sawtooth in addition to sine (or even just triangle) 
oscillators that can be offset (and ideally modulated) individually, 
I'd say eight oscillators is almost overkill already.

> Another thing that popped up is: How "correct" do they have to be?

If you want to create waveforms, they have to be very precise and in 
particular the need to be phase locked. In any case if you want to 
experiment with precision harmonics, get a K5000 (a subset of additive) 
or an XT (wavetable, which is equivalent to almost the same subset of 
additive).

The appeal of the additive VCO you proposed is the implied 
imperfection, so forget about the waveform whose spectrum you're trying 
to recreate and just dial up the main harmonics - it may not look 
anything like that waveform on the scope, but it sure sounds like it. 
You have to keep in mind that no acoustical instrument produces 
anywhere near a perfect harmonic series (strings produce stretch from 
being stiff etc.pp.). The sound of wavetable and some additive synths 
which have perfect harmonics by construction sound so distinctively 
"artificial" because of that property IMHO.

On Monday 08 March 2004 06:43, John L Marshall wrote:
> Another approach would be to derive sine paritals from a Walsh
> generator.

A Walsh generator needs a good deal of filtering to get a clean sine. 
It is probably cheaper to just store the sequence than to produce it 
live as well, at least you can quite cheaply produce partials from the 
same sequence. However there's still that multibit D/A converter at the 
end, but there are "magic binary sequences" that need just to be 
integrated to produce a sine to good approximation. Sort of a 
precomputed Delta-Sigma converter... I can't seem to find the link I 
had for some of these sequences, but it was buried in some stepping 
motor control treatise IIRC. Anyway, all you need is a shift-register 
pre-loaded with the correct sequence and a simple integrator for D/A. 
If you can get hold of video RAM or FIFO SRAM, you get eight or sixteen 
phase-locked and harmonically perfect generators. Loading in different 
sequences gives different waveforms, too. And since the sequence is 
probably much longer than you need, you can even modulate the wave 
within the sequence already. Or you go back to multibit conversion and 
have an even longer sequence (or multiple sequences) available.


Achim.
-- 
+<[Q+ * Matrix-12 * WAVE#46 * microQkb/Omega * XTk/30 * sonic heaven]>+

SD adaptation for Waldorf rackAttack V1.04R1:
http://homepages.compuserve.de/Stromeko#WaldorfSDada