[sdiy] Low Frequency Square to Sine Waveshaper

[Magnus Danielson]

From: "Paul Perry" <pfperry at melbpc.org.au>
Subject: RE: [sdiy] Low Frequency Square to Sine Waveshaper
Date: Tue, 27 May 2008 00:51:30 +1000
Message-ID: <009c01c8bf3f$fc2c84c0$8501a8c0 at ice9>

> I'd agree that the idea of a high frequency clock divided down & driving a
> kind of sine construction - the approach made famous by Don Lancaseter, eg
> at www.tinaja.com/glib/stepsynt.pdf is very good indeed for audio
> applications.
> 
> But is there a problem when it comes to LFOs? When a very low frequency LFO
> is modulating an oscillator, a discontinuity that would be quite
> unnoticeable in an audio signal, is only too obvious. LFOs are definitely a
> special - and very awkward - case.
> 
> That is why I think a phase locked method is more promising. Although, even
> there it won't be without problems, when if comes to settling times in the
> PLL. At LFO frequencies, the settling time might be embarrassingly long.
> 
> The more I think about it, the better triangles look.

I agree with you that you do need to care when it comes to LFOs, but it is not
important that it is perfectly sine, but has the same smoothness. My 5 bit
proposal was maybe running a bit short, but also has a hidden problem to it.
It won't work as well as I thought... I did not properly think about the
overtone spectras. Supprised nobody spotted the error. You can't just cancel
out the unwanted odd overtones with even overtones. Just wont cut it.

You can generate arbitrary triangles and bend those into sine-like creatures.

Cheers,
Magnus