From: Clay on
On Mar 18, 6:23 pm, robert bristow-johnson <r...(a)audioimagination.com>
wrote:
> On Mar 18, 11:52 am, dbd <d...(a)ieee.org> wrote:
>
>
>
>
>
> > On Mar 18, 6:09 am, "Junglist" <vasily.karpenko(a)n_o_s_p_a_m.gmail.com>
> > wrote:
>
> > > Hello!
>
> > > I have read article "Optimum Masking Levels and Coefficient Sparseness for
> > > Hilbert Transformers and Half-Band Filters Designed Using the
> > > Frequency-Response Masking Technique", Yong Ching Lim, NOVEMBER 2005.
>
> > > There're in example two filters Hb(z) and H1(z). I guess they derived by
> > > multiplication classical hilbert impulse function h(n)=[1-cos(Pi*n)]/Pi*n
> > > on different windows. What kind of windows is using there?
> ..
> > Why do you suggest the use of windows here? The frequency response
> > masking literature takes advantage of a variety of filter design
> > methods, but usually optimizing techniques.
>
> an implied window can come from any design technique as long as you
> can avoid dividing a non-zero numerator by a zero denominator.
>
> because half-band symmetry let's us ditch the even-numbered taps, any
> design that imposes half-band symmetry can have its (properly aligned)
> impulse response divided by the ideal
>
>    h[n] = (1 - (-1)^n)/(pi*n)        (h[0]=0)
>
> for odd n, and you have an implied window.
>
> r b-j- Hide quoted text -
>
> - Show quoted text -

True, but the article refers to Chebyshev approximation and the effect
the masking has on its ripple, so I assume he's using a Remez method
to obtain his original filters. And then "sharpening" them from there.

My 2 cents worth anyway.

Clay
From: robert bristow-johnson on
On Mar 19, 10:29 am, Clay <c...(a)claysturner.com> wrote:
> On Mar 18, 6:23 pm, robert bristow-johnson <r...(a)audioimagination.com>
> wrote:
>
>
>
> > On Mar 18, 11:52 am, dbd <d...(a)ieee.org> wrote:
>
> > > On Mar 18, 6:09 am, "Junglist" <vasily.karpenko(a)n_o_s_p_a_m.gmail.com>
> > > wrote:
>
> > > > Hello!
>
> > > > I have read article "Optimum Masking Levels and Coefficient Sparseness for
> > > > Hilbert Transformers and Half-Band Filters Designed Using the
> > > > Frequency-Response Masking Technique", Yong Ching Lim, NOVEMBER 2005.
>
> > > > There're in example two filters Hb(z) and H1(z). I guess they derived by
> > > > multiplication classical hilbert impulse function h(n)=[1-cos(Pi*n)]/Pi*n
> > > > on different windows. What kind of windows is using there?
> > ..
> > > Why do you suggest the use of windows here? The frequency response
> > > masking literature takes advantage of a variety of filter design
> > > methods, but usually optimizing techniques.
>
> > an implied window can come from any design technique as long as you
> > can avoid dividing a non-zero numerator by a zero denominator.
>
> > because half-band symmetry let's us ditch the even-numbered taps, any
> > design that imposes half-band symmetry can have its (properly aligned)
> > impulse response divided by the ideal
>
> >    h[n] = (1 - (-1)^n)/(pi*n)        (h[0]=0)
>
> > for odd n, and you have an implied window.
>
....
>
> True, but the article refers to Chebyshev approximation and the effect
> the masking has on its ripple, so I assume he's using a Remez method
> to obtain his original filters. And then "sharpening" them from there.

still, an implied window can be derived from the data as long as there
are no 1/0 kind of division. even when using Parks-McClellan, you can
enforce half-band symmetry, which will make the even samples zero.
then the conditions are met and an implied window can be observed.

r b-j

From: Clay on
On Mar 20, 5:06 pm, robert bristow-johnson <r...(a)audioimagination.com>
wrote:
> On Mar 19, 10:29 am, Clay <c...(a)claysturner.com> wrote:
>
>
>
>
>
> > On Mar 18, 6:23 pm, robert bristow-johnson <r...(a)audioimagination.com>
> > wrote:
>
> > > On Mar 18, 11:52 am, dbd <d...(a)ieee.org> wrote:
>
> > > > On Mar 18, 6:09 am, "Junglist" <vasily.karpenko(a)n_o_s_p_a_m.gmail..com>
> > > > wrote:
>
> > > > > Hello!
>
> > > > > I have read article "Optimum Masking Levels and Coefficient Sparseness for
> > > > > Hilbert Transformers and Half-Band Filters Designed Using the
> > > > > Frequency-Response Masking Technique", Yong Ching Lim, NOVEMBER 2005.
>
> > > > > There're in example two filters Hb(z) and H1(z). I guess they derived by
> > > > > multiplication classical hilbert impulse function h(n)=[1-cos(Pi*n)]/Pi*n
> > > > > on different windows. What kind of windows is using there?
> > > ..
> > > > Why do you suggest the use of windows here? The frequency response
> > > > masking literature takes advantage of a variety of filter design
> > > > methods, but usually optimizing techniques.
>
> > > an implied window can come from any design technique as long as you
> > > can avoid dividing a non-zero numerator by a zero denominator.
>
> > > because half-band symmetry let's us ditch the even-numbered taps, any
> > > design that imposes half-band symmetry can have its (properly aligned)
> > > impulse response divided by the ideal
>
> > >    h[n] = (1 - (-1)^n)/(pi*n)        (h[0]=0)
>
> > > for odd n, and you have an implied window.
>
> ...
>
> > True, but the article refers to Chebyshev approximation and the effect
> > the masking has on its ripple, so I assume he's using a Remez method
> > to obtain his original filters. And then "sharpening" them from there.
>
> still, an implied window can be derived from the data as long as there
> are no 1/0 kind of division.  even when using Parks-McClellan, you can
> enforce half-band symmetry, which will make the even samples zero.
> then the conditions are met and an implied window can be observed.
>
> r b-j- Hide quoted text -
>
> - Show quoted text -

I wasn't saying you can't do it this way, but rather I was reflecting
on the OP's question about what window or how the particular filters
in the article were created.

Clay