Prev: On Quaternions -- Book Recommendations?
Next: Practical Bit-loading (ADSL/VDSL): fundamental problem
From: jungledmnc on 7 May 2010 16:17
Hi,

I have implemented overlap-add method for granular pitchshifting and
eventually I realized, that since the "overlapping" is basically
multiplying with a hann window or something, it basically is a
ring-modulation. I unfortunately verified this on a spectrum analyzer - it
added f+m and f-m to the testing frequency. But this means the only way to
avoid the ringing is to make the window larger and generally you can never
get rid of it completely. Is that correct or am I missing something?

Thanks!
From: Dirk Bell on 7 May 2010 18:36
On May 7, 4:17 pm, "jungledmnc" <jungledmnc(a)n_o_s_p_a_m.gmail.com>
wrote:
> Hi,
>
> I have implemented overlap-add method for granular pitchshifting and
> eventually I realized, that since the "overlapping" is basically
> multiplying with a hann window or something, it basically is a
> ring-modulation. I unfortunately verified this on a spectrum analyzer - it
> added f+m and f-m to the testing frequency. But this means the only way to
> avoid the ringing is to make the window larger and generally you can never
> get rid of it completely. Is that correct or am I missing something?
>
> Thanks!

Are you asking if the overlap and add process with a suitable window
(long or short), but no other processing causes bandwidth expansion in
the final result? NO.

Are you asking if the overlap and add process with a suitable window,
but no other processing causes bandwidth expansion in the intermediate
results (i.e. individual frames before they are added)? YES.

Are you asking if something in the processing you are doing causes
bandwidth expansion in the final result? PROBABLY.

With a suitable window windowing, in individual frames you will see a
bandwidth expansion, but in some applications it is possible to have
that effect cancel out when the frames are recombined. You may not be
doing one of those applications, or may not be doing it in such a way
as to have the desired cancellation (may not be possible).

Dirk


Dirk
 | 
Pages: 1
Prev: On Quaternions -- Book Recommendations?
Next: Practical Bit-loading (ADSL/VDSL): fundamental problem