[sdiy] VCO design question

[Tom Wiltshire]

On 2 Dec 2008, at 20:16, Antti Huovilainen wrote:

> On Tue, 2 Dec 2008, Paul Maddox wrote:
>
>> I wouldn't say any idea was bad, just because it doesn't follow  
>> the rules :-)
>
> For a valid comparison it has to have same quality and features as  
> the analog solution though. Apples to apples and all that.
> Digital oscillator can do that, but it's no longer quite trivial.
>
> You can "pulse width" modulate most waveforms if you synthesize two  
> copies of the integral of said waveform and adjust their phase  
> difference (sawtooth is integral of pulse train).

Sorry, but I can't help myself but jump in here, even if is going  
rather off topic. Poor David'll never get his answer...

I've recently done this ramp down/triangle/ramp up thing on a digital  
osc. I didn't use the integral technnique that Antti mentions, but  
instead used a simple piecewise linear phase distortion function. I  
did it principally because I wanted the oscillator to have proper  
PWM, but the "digital bonus" is that you can apply the effect to all  
the other waveforms too, not just square waves.
It wasn't particularly difficult (around 15-20 instructions on a  
dsPIC) but it isn't trivial. Even 20 instructions starts to add up  
between samples.

The other problem is that "squashing" a triangle wave into a ramp  
wave increases the harmonic content. This happens with any other wave  
too, even a sine. So aliasing is an issue at higher frequencies. How  
much of a problem obviously depends on the source waveform and  
frequency, and the sampling frequency. I found that I needed to limit  
PWM to the range 5%-95% to avoid the worst effects, but this is  
subjective. Probably others would go to 2%-98%, or even revel in the  
nasty crunchiness of it all.

T.


.