Re: AW: Re: aliasing again

From Magnus Danielson
Sent Fri, Oct 18th 1996, 20:09

> > When you do a DAC thing you will have the spectrum being folded up several
> > times. So to suppress the mirror waveforms you _will_ need an filter on 
> the
> > output.
> [...]
> >So the mirrors needs to be cut out by a rather steep low pass filter.
> Yes, you're right. Good description btw.!

Thank you!

> My original point was that you can't prevent aliasing with a filter after
> the dac (i.e. if the mirrored spectra already overlap, you're lost).
> Just forgot that you still have to filter the mirrors out, not just the
> clock, in the general case - sorry.

A well.... I think that I understand your original point.
You need to filter the input to avoid haveing the spectra 'enhanced' with 
frequencies mirrored down, those will be back in place througth mirror but
spread all over...

> What I had in mind was a special case, when the playback clock is
> some HF-VCO, which directly reads out the date from a ram.
> I think the sample module of the old doepfer modular
> worked this way, and some drum machines. This approach should
> make the spectral lines of the mirrors harmonics of the original as well.
> Lack of filter just makes the sound brighter, but not unharmonic.

Interesting type of design that I also have considerd. A way to improve things
would be to oversample the VCO and use a VCF to filter out the mirrors.