[sdiy] What is a quadrature oscillator good for?

[jhaible at debitel.net]

>    > http://www.oldcrows.net/~jhaible/tonline_stuff/hjfs2_qv.gif
> 
> I had considered this approach, but didn't use it for two reasons:
> 
> One is that the bipolar control voltage splits into an absolute value
> current and a polarity, and any offset voltage in that area will show
> up as an imperfect transition across zero Hz.

Right. And this is why your design is so attractive for me as well!


> The other is that any offset voltages in the integrators will cause
> the triangle outputs to slew at wildly different rates at frequencies
> near zero Hz.  Possibly even slewing in different directions.

I don't think so. There is this 200k feedback resistor from the 2nd
integrator to the schmitt trigger that will (hopefully) prevent this.
With "ideal" components, this would not be needed (just one feedback
from 1st integrator to schmitt trigger): A perfect integrator would
be enough to get a 90deg triangle by integrating sign(1st triangle).
Further, an ideal integrator would not interfere with the schmitt 
trigger action (as controlled from the first triangle), because 
the 2nd triangle is exactly zero when the first triangle reaches
the trigger points.
Now, with a real life integrator (offset currents and all), and
slight imbalance of the 2nd triangle (not being zero when first
triangle reaches nominal trigger points) will slightly change
the trigger points, and thus correct any offset in the 2nd 
integrator. (It's quite fascinating how this is achieved with
a single resistor, IMO.)



>    > I haven't checked this in theory, but in all these years that I
>    > have this Frequency Shifter, the VCO _never_ got stuck anywhere.
> 
> Perhaps an offset voltage generating the control currents keeps the
> stuck state from occurring.  That's a wild guess.

I don't think it's offset currents. 

Maybe it just didn't occur yet, because I have so few thru-zero
transitions (even in several years) because I don't use it for
audio rate modulation. Who knows.

JH.

-------------------------------------------------
debitel.net Webmail