Constrained Least Square FIR Filter Design - Docementation


Go back to previous page.

Programs for constrained L2 linear phase FIR filter design.

The integral square error is minimized subject to a
constraint on the size of the peak error size in the
passband and stopband.  When a weighting function is not
specified, the unweighted square error over [0,pi] is
minimized.

____________________________________________________________
bare bones (simple, unadorned programs)_____________________

cl2lp.m
        constrained L2 filter design with out specified band
        edges.
          - lowpass only
          - odd length only

help on cl2lp.m:
        h = cl2lp(m,wo,up,lo,L)
        h  : 2*m+1 filter coefficients
        m  : degree of cosine polynomial
        wo : cut-off frequency in (0,pi)
        up : [upper bound in passband, stopband]
        lo : [lower bound in passband, stopband]
        L  : grid size
        example
        up = [1.02, 0.02]; lo = [0.98, -0.02];
        h = cl2lp(30,0.3*pi,up,lo,2^11);

cwl2lp.m
        constrained weighted L2 filter design.
          - lowpass only
          - odd length only
          - the weighting function has specfied band edges
          - the constraints do not have specfied band edges

adamsfir.m
        program for constrained L2 filter design based on 
        Adams' quadratic programming formulation.
          - lowpass only
          - odd length only
          - both the L2 weighting function and the 
            constraints have specified bands edges

cl2bp.m
        constrained L2 bandpass filter design with out 
        specified band edges.
          - bandpass only
          - odd length only

____________________________________________________________
More programs ______________________________________________

cl2lpr.m
        constrained L2 filter design with out specified band
        edges.
          - lowpass only
          - uses Newton's method to refine extremal 
            frequencies
          - handles both odd and even lengths
          - returns several parameters

cwl2lpr.m
        constrained weighted L2 filter design.
          - lowpass only
          - uses Newton's method to refine extremal
            frequencies
          - handles both odd and even lengths
          - returns several parameters
          - the weighting function has specfied band edges
          - the constraints do not have specfied band edges

cl2mbr.m
        constrained L2 multiband filter design with out 
        specified band edges.
          - uses Newton's method to refine extremal 
            frequencies
          - handles both odd and even lengths
          - returns several parameters

adamslpr.m
        Adams approach to constrained L2 filter design.
          - uses Newton's method to refine extremal 
            frequencies
          - handles both odd and even lengths
          - returns several parameters
          - both the L2 weighting function and the 
            constraints have specified bandedges.

____________________________________________________________
More programs ______________________________________________

fircls1.m
        constrained L2 lowpass and highpass filter design 
        with out specified band edges.
          - uses Newton's method to refine extremal 
            frequencies
          - handles both odd and even lengths
          - returns only the filter
          - permits the user to specify a bandedge: (either 
            passband or stopband, but not both)

fircwls1.m
        same as fircls1.m, but has nonuniform L2 weighting 
        capability.

fircls.m
        constrained L2 filter design with out specified 
        band edges.
          - works for multiband filter types
          - uses Newton's method to refine extremal 
            frequencies
          - handles both even and odd lengths
          - returns only the filter

____________________________________________________________
subprograms provided: ______________________________________

local_max.m
frefine.m
ffbe.m
____________________________________________________________
      

Please report any bugs or send comments regarding the programs to selesi@ece.rice.edu