Prev: filtering in frequency domain
Next: A very basic question about DSK
From: Fred Marshall on 5 Jul 2010 13:00
Long ago I wrote a Remez exchange program that accepted a continuous
magnitude response specification rather than the P-M fixed band type of
specification.

Rather than excluding the transition bands, it included them and this
makes using it a bit trickier. In that sense P-M is better. Yet, it's
sometimes nice to be able to specify the magnitude response as other
than a set of flat or straight lines.

Of course, the two could be combined so that the critical passband
(usually not stopband) shapes could be specified and the transition
bands could be left out.

Ditto on the weighting functions.

Does anyone know of a modern program that will allow this type of
specification?

Fred
From: robert bristow-johnson on 5 Jul 2010 14:27
On Jul 5, 1:00 pm, Fred Marshall <fmarshallx(a)remove_the_xacm.org>
wrote:
> Long ago I wrote a Remez exchange program that accepted a continuous
> magnitude response specification rather than the P-M fixed band type of
> specification.
>
> Rather than excluding the transition bands, it included them and this
> makes using it a bit trickier.  In that sense P-M is better.  Yet, it's
> sometimes nice to be able to specify the magnitude response as other
> than a set of flat or straight lines.
>
> Of course, the two could be combined so that the critical passband
> (usually not stopband) shapes could be specified and the transition
> bands could be left out.
>
> Ditto on the weighting functions.
>
> Does anyone know of a modern program that will allow this type of
> specification?

i've used MATLAB's firpm() (formerly remez()) so that the spec (both
target and weights) was a sorta large set of little broken lines that
could fit a given specified function. it worked good enough for me
that i hadn't felt motivated to try to redo firpm() from scratch.
wasn't a big deel.

r b-j


 | 
Pages: 1
Prev: filtering in frequency domain
Next: A very basic question about DSK