The process does not know anything. It's just that
a digital sinusoid at say sr/2, is basically [1, -1, 1,
-1,..].
When this is decoded by the DAC, the analogue wave
is a sinusoid with all the 'in-between' values.

The problem with anything above sr/2 is that it cannot
be represented, or is mis-represented by a sinusoid below
sr/2. Note that a sinusoid at sr/2 is not aliased, but is
critically
sampled.

Non-bandlimited sawtooths and square waves (and any other
types of wave) will have components above sr/2, so they
cannot
be properly represented in digital form.

Victor
>
>   Victor Lazzarini  wrote
>
> >  When it is played back the ADC process lowpass filters
> > the signal and the squarish wave becomes smooth.
>
> Then how would this process "know" when it's a saw tooth
> wave that should stay rough ? Is there a predominance of
> sine waves over the other kinds, as Luis Jure understands:
>
> > if you consider that a signal with a complex waveform
> > can be  represented by
> > a sum of sinusoids, and you have a waveform with spikes,
> > angles or  anything
> > that means a discontinuity in the derivative, think that
> > you'll need an infinite series of sinusoids to
> approximate that form.
>
> Why a square wave does not produce aliasing?
>
> --
> Send bugs reports to this list.
> To unsubscribe, send email to
> csound-unsubscribe@lists.bath.ac.uk