[sdiy] BLIT/BLEP virtual analogue synthesis

[Olivier Gillet]

> I'm not sure I understand what you are trying to get at.  A sigma-delta
> modulator, much simplified, takes a slow signal of high resolution and
> converts it into a fast signal of low resolution.  The frequency content of
> the modulated signal is massaged so that the much increased quantization
> noise ends up at high frequencies with lots of space to the frequency band
> of interest, so it can be easily filtered during signal reconstruction.  The
> other virtue of the low-resolution signal is that it is easier to have a
> linear conversion to analog.
> You normally wouldn't do that, but as long as the high-frequency content is
> noise-like (not overtly tonal and of much less amplitude than the signal of
> interest) it will probably just raise the noise floor.  Higher-order loop
> topologies on the other hand might not be stable in that case.

Let's take the case of a sigma-delta loop operating at 8 MHz. In the
"typical" use case, the audio input at 48kHz is oversampled by 166, so
only 1 / 166 of the bandwidth is occupied.
We could also feed it data directly computed at 8 MHz (like a brute
force oscillator), in which case 100% of the bandwidth, up to 4 MHz
will be occupied. Will this be a source of problems? Or will this only
create more high-frequency junk to filter?

I guess I'll just fire a python shell and find out...