AW: quadrature osc

[Martin Czech]

> 	>Yesterday I grabbed into a book shelf and found an old Elektor
> Audio
> 	>Special: Low distortion sine generator. The idea was to set up a
> second
> 	>order differential equation (ie. state var filter). It looks very
> 	>much the same then J.H.'s triangle quadrature oscillator. 
> 
> It looks similar at the first glance, but it's actually very different.
> The second order differential equation uses a continuous voltage
> to drive the integrators, thus all the amplitude problems you 
> described. My oscillator (and the Tietze/Schenk circuit from
> which I've derived it) uses comparators to drive the integrators.
> There is also a regulation path (not for amplitude, but for offset
> voltage of the 2nd integrator), but the trick is that this path will be
> completely inactive when the initial error is corrected. No time
> constants involved.

I know, I know, but I mean , the cicuit topology looks quite the
same. It is just nonlinear, and the comparator enshures amplitude
regulation.

> With the 2nd order differential equation method, you must make 
> a compromise between response time of your regulation, and
> waveform distortion. This is easy for high VCO frequencies, but
> almost impossible for LFO applications.
> The cos**2 + sin**2 method you mentioned is the one exception
> to this dilemma. No time constants involved here, too.
> I'd be interested in results from somebody who has tried it.

With the strange hybrid chip Elektor claimed 0.01% THD (from
my head). This is why I ask all these stupid questions about
dome filters and their errors, quadrature osc. and their errors etc,
because I'm interested in good sideband rejection. These
errors will contribute.

Will the state variable also give through zero modulation.  From my
head I think there's no reason why this shouldn't be.  The integration
principle is the same, also the influence of "right" comparator direction
on stability.

And: I have the feeling that quadrature error is less for the state var at
higher frequencys, because of lower neccessary bandwidth.  No harmonics.

Maybe THD is not so important, the CA3080 waveshaper has certainly about
a few %, so a nonlinear device in fedback path may be sufficient. A
resistor and a diode bridge as in the ETI2000 (now I finally understand
what this bridge is doing there !). In a state var, a lower feedback
resistance makes MORE damping.

Just a few stupid ideas.

m.c.