Hi Raj,
  I am guessing that you are talking about a technique were you insert
zeros inbetween the filter taps to generate a "periodic" filter, then
you create another filter "the masking filter" with a lower complexity
to get the passband of interest.  I've seen this technique used to
generate narrowband filters used for decimation.  Do a google search
on "frequency response masking" and you will find much information on
the technique.

Jon 

raaj_cee@msn.com (Rajasekaran C) wrote in message news:<6e6d57f7.0309250151.6cd57523@posting.google.com>...
> Hi,
>    Can anyone please explain how decimation using frequency response
> masking is done?? thanks in advance.
> 
> Raj.

Rajasekaran C wrote:

> Hi,
>   Thanks for the reply. What I meant by "frequency response masking"
> was the removal of the frequency components above the new sample rate
> by using masking filters. i.e., the unwanted frequencies are masked
> using filters. I wanted to know whether a decimation filter can be
> implemented using this technique. Thanks once again for the reply.
>=20
> Regards,
> Raj.

Raj,

I guessed your meaning correctly. The process of decimating by an
integer ratio 1/N amounts simply to selecting every Nth sample. As you
wrote, frequencies that the new sample rate won't support need to be
filtered out first. The operations can be combined by using a
transversal filter (the usual kind of FIR) and calculating only every
Nth output value. A slight embellishment can be used for ratios like
M/N. You might want to look at polyphase approach exemplified by
http://www.nt.ruhr-uni-bochum.de/lehrstuhl/dsv/publikationen/ecctd99_1.pd=
f

Jerry
--=20
Engineering is the art of making what you want from things you can get.
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF

Hi,
  Thanks for the reply. What I meant by "frequency response masking"
was the removal of the frequency components above the new sample rate
by using masking filters. i.e., the unwanted frequencies are masked
using filters. I wanted to know whether a decimation filter can be
implemented using this technique. Thanks once again for the reply.

Regards,
Raj.

Jerry Avins <jya@ieee.org> wrote in message news:<bkuv53$o9d$1@bob.news.rcn.net>...
> Rajasekaran C wrote:
> 
> > Hi,
> >    Can anyone please explain how decimation using frequency response
> > masking is done?? thanks in advance.
> > 
> > Raj.
> 
> Raj,
> 
> I have to guess at your meaning. If you're using a standard term, it
> refers to something I don't know about. If, by "frequency response
> masking", you mean what I call filtering, then the answer to your 
> question is simple, even if my explanation is not.
> 
> When decimating a signal -- reducing its sample rate -- the maximum 
> supportable frequency is also reduces. That part of the original signal 
> that lies above half the new sample rate must be reduced (to avoid
> aliasing in the decimated signal) before samples are discarded.
> 
> Those filter outputs that are to be discarded need not be calculated at
> all unless the workings of a recursive filter needs them. The actual
> procedure that is best will depend on the degree of decimation, which
> you didn't' tell us.
> 
> Jerry

Rajasekaran C wrote:

> Hi,
>    Can anyone please explain how decimation using frequency response
> masking is done?? thanks in advance.
>=20
> Raj.

Raj,

I have to guess at your meaning. If you're using a standard term, it
refers to something I don't know about. If, by "frequency response
masking", you mean what I call filtering, then the answer to your=20
question is simple, even if my explanation is not.

When decimating a signal -- reducing its sample rate -- the maximum=20
supportable frequency is also reduces. That part of the original signal=20
that lies above half the new sample rate must be reduced (to avoid
aliasing in the decimated signal) before samples are discarded.

Those filter outputs that are to be discarded need not be calculated at
all unless the workings of a recursive filter needs them. The actual
procedure that is best will depend on the degree of decimation, which
you didn't' tell us.

Jerry
--=20
Engineering is the art of making what you want from things you can get.
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF

Hi,
   Can anyone please explain how decimation using frequency response
masking is done?? thanks in advance.

Raj.