[sdiy] square to sine

[Richie Burnett]

> On the other hand, why couldn't I simply AC-couple the square output to 
> the
> integrator input?  This might cause a rounding of the triangle apices, but
> I'm going to shape it into a sine anyway, so does it matter?  Of course,
> this doesn't address the integrator offset, does it?

AC coupling the square wave to the integrator will result in some "LF droop" 
on the portions of the square-wave that should otherwise be flat.  Once 
integrated this will degrade the linearity of the resulting triangle.  If 
you plan on wave-shaping the triangle into a sinewave then any triangle 
asymmetry will directly cause sinewave asymmetry and increase THD.

> The other alternative is to use a lossy integrator...(thinking aloud)...?

A lossy integrator is really a 1st-order low-pass filter.  So in theory it 
is equivalent to the method you already described above.

You have a first-order high-pass filter followed by a pure integrator,  _OR_ 
first-order low-pass filter in the form of a leaky integrator.

The zero at DC for the high-pass filter in the first case would cancel the 
pole at DC for the integrator leaving just a single pole at the cutoff 
frequency of the high-pass filter.  This is the same thing as the single 
pole that results from breaking a pure integrator's gain back to flat at 
low-frequencies in the form of a lossy integrator.  (The 1/s term for the 
integrator effectively converts the 1st-order HPF into a 1st-order LPF.)

-Richie,