ada@munsell.UUCP (Andy Adams) (03/27/91)
I'm currently involved in FIR digital filter design.
I've been using the CARL digital filter design software
by Mcclellan and Parks designed some 10 years ago.
I am trying to design fairly large (16->32 tap) FIR
filters for image sharpness correction/enhancement. Ideally,
I would like to be able to specify filter gains at various
frequencies and have the program optimally design the coeffs.
The band-edge, band-gain, band-ripple approach of CARL
seems to be very awkward and unpredictable for this
type of filter design.
I'm also designing very large (256->1024) tap interpolation
filters. CARL is limited to 128 coeffs.
If anybody knows of any more recent offerings or newer
algorithms for filter design/analysis software, please
post or e-mail some recommendations.
I've seen some references to Gabriel from UC Berkley.
Is this any good? How do I obtain it? Does it run on a Sun?
Can I ftp something??
I realize this has probably been asked in the past, but
I haven't captured the info.
I would prefer that to run on Sun workstation (Sun3 or Sparc).
Thanks much,
-----------------------------------------------------------------------
Andy Adams, KEPS Inc. |internet: ada@epps.kodak.com
an Eastman Kodak Company |
164 Lexington Rd. |
Billerica, Ma. 01821 |
(508)670-6593 |
-----------------------------------------------------------------------hanke@nessie.cs.id.ethz.ch (Norbert Hanke) (03/28/91)
In article <2345@gonzales.munsell.UUCP> ada@munsell.UUCP (Andy Adams) writes: > > I am trying to design fairly large (16->32 tap) FIR > filters for image sharpness correction/enhancement. Ideally, > I would like to be able to specify filter gains at various > frequencies and have the program optimally design the coeffs. > The band-edge, band-gain, band-ripple approach of CARL > seems to be very awkward and unpredictable for this > type of filter design. Try to use them Remez Exchange Algorithm, if you want to design linear phase filters. Internally, Remez builds a large grid of the specified frequencies, each point having its desired value and the weight. It should be possible to adapt Remez in a way to allow for arbitrary frequency responses, not just the band-edge, band-gain, band-weight approach. Norbert Hanke Power Electronics & Electrometrology ETH Zurich, Switzerland