[sdiy] VA vs. RA: Square wave

[Paul Schreiber]

> I have noticed in some analogue oscillators squarewave output, that
> there is an unchanging time constant governing the loss of voltage
> from the "high" state of the oscillator, such that at low frequencies
> you actually see a gentle ramp downward in voltage from the start of
> the cycle until the next reset, instead of a perfect square.  Since
> this ramp downwards is not frequency-dependent (or not so much), at
> high frequencies the same oscillator puts out a more perfect square
> wave.

The VCOs are not doing this. It's a function of inter-stage AC coupling.

The caps (and the associated input impedance) create a quasi-BP filter (mostly 
HP). In order not to see this on a scope, the cutoff needs to be about 5 times 
below the lowest frequency. This can lead to rather large caps (over 100uf) 
which are physically large and expensive. So, a good compromise is to use 10uf 
caps (they need to be non-polar). Assuming a 33K input impedance for an audio 
summer, that gives a ~0.5HZ cutoff so anything above say 30Hz would not exhibit 
much "tilt".

Where you really start to notice it is when you use a AC coupling cap of say 
0.47uf or below.

For a VA, this can be accounted for by either tilting the VCO output itself 
(*all* waveforms are effected, it's just easier to see it with a square) or a 
simple IIR filter to emulate the coupling effects.

Paul S.