[music-dsp] Re: Declipping

Yaakov Stein yaakov_s at rad.com
Wed Feb 23 04:01:05 EST 2005


 
> Hm, "just wrong"--pretty heavy ;-)

I agree with Bob.

Assuming there was an original BW analog signal
properly sampled, the doubling of the sampling rate
should leave the spectrum would need to leave the
spectrum the same.

In fact, adding a zero between every two samples
adds spectral components in the higher half,
which should have been empty. These need
to be filtered out to regain the original spectrum,
which in the time domain puts the properly
interpolated values in between the samples.

Of course, had the original analog signal really had
zeros between every low-rate samples 
(e.g. have been at least twice the original BW),
then it would have been quite a different signal!
In fact, we could have by chance sampled it
at the other phase, resulting in a completely
zero signal. This is clearly not the same signal
as the one in question.

So the zeros we insert are completely artificial.
In fact, you could insert any other constant value,
such as your maximum value, with roughly the same effect.

Jonathan (Y) Stein




More information about the music-dsp mailing list