[music-dsp] freq. shifter

[eli+ at gs211.sp.cs.cmu.edu]

Sampo Syreeni wrote:
> On Mon, 11 Mar 2002, Laurent de Soras wrote:
> >It works well when the modulating frequency is greater than half the
> >higher frequency of the modulated signal ; when the lower aliased
> >sideband doesn't go over the upper sideband.

Modulating frequency should be the frequency of the top of the band
you're inverting, I'd think, and it's okay if the signal fills almost
the whole band.

> Actually it will always do that -- the next period of the spectrum will
> descend over the modulated version as soon as you start your processing.

Hmm, periodic spectrum?  A sampled signal's spectrum?

> In a digital implementation the gimmick is, when you modulate precisely at
> the Nyquist frequency, the different periods of the spectrum will align
> perfectly, in phase.
> 
> I reallly don't know what it is that makes the analog scramblers
> invertible.

Or are you referring to when you ring-mod to invert AB, you get BAAB,
and when you try to do it again you get ABBAAB but the bottom
(in-band) AB is doubled?  So the two mod oscs have to be phase-locked
or your signal beats slowly in and out.

I don't know what the scramblers did, but you could build a lowpass
into the inverter, cutting off above the band.  So you get BAab after
one iteration, and attenuated beating after two.  (If you had a
general SSB modulator lying around you could use that (shift down
only), but it's overkill here where the two sidebands don't overlap.)

-- 
     Eli Brandt  |  eli+ at cs.cmu.edu  |  http://www.cs.cmu.edu/~eli/

dupswapdrop -- the music-dsp mailing list and website: subscription info,
FAQ, source code archive, list archive, book reviews, dsp links
http://shoko.calarts.edu/musicdsp/