PDL::DSP::Simple - Simple interface to windowed sinc filters.


       use PDL::LiteF;
       use PDL::DSP::Fir::Simple;


At present, this module provides one filtering routine. The main purpose is to provide an easy-to-use lowpass filter that only requires the user to provide the data and the cutoff frequency. However, the routines take options to give the user more control over the filtering. The module implements the filters via convolution with a kernel representing a finite impulse response function, either directly or with fft. The filter kernel is constructed from windowed sinc functions. Available filters are lowpass, highpass, bandpass, and bandreject. All window functions in PDL::DSP::Windows are available.

See "moving_average" in PDL::DSP::Iir for a moving average filter.

Some of this functionality is already available in the PDL core. The modules PDL::Audio and PDL::Stats:TS (time series) also have filtering functions.

Below, the word order refers to the number of elements in the filter kernel. The default value is equal to the number of elements in the data to be filtered.

No functions are exported by default.



  $xf = filter($x, {OPTIONS});


  $xf = filter($x, $kern);


Apply lowpass filter to signal $x with a cutoff frequency of 90% of the Nyquist frequency (i.e. 45% of the sample frequency).

 $xf = filter($x, { fc => 0.9 });

Apply a highpass filter rather than the default lowpass filter

  $xf = filter($x, {fc => 0.9 , type => 'highpass' });

Apply a lowpass filter of order 20 with a blackman window, rather than the default hamming window.

  $xf = filter($x, {fc => 0.9 , window => 'blackman' , N => 20 });

Apply a 10 point moving average. Note that this moving averaging is implemented via convolution. This is a relatively inefficient implementation.

  $xf = filter($x, {window => 'rectangular', type => 'window', N => 10 });

Return the kernel used in the convolution.

  ($xf, $kern)  = filter($x, { fc => 0.9 });

Apply a lowpass filter of order 20 with a tukey window with parameter alpha = 0.5.

  $xf = filter($x, {fc => 0.9 , 
    window => { name => 'tukey', params => 0.5 } , N => 20 });



Order of filter. I.e. the number of points in the filter kernel. If this option is not given, or is undefined, or false, or less than zero, then the order of the filter is equal to the number of points in the data $x.


A kernel to use for convolution rather than calculating a kernel from other parameters.


Boundary condition passed to convolveND. Must be one of 'extend', 'truncate', 'periodic'. See PDL::ImageND.

All other options to filter are passed to the function "firwin" in PDL::DSP::Fir which creates the filter kernel. "firwin" in PDL::DSP::Fir in turn passes options to PDL::DSP::Windows:window. The option window is either a string giving the name of the window function, or an anonymous hash of options to pass to PDL::DSP::Windows:window. For example { name => 'window_name', ... }.

If the second argument is not a hash of options then it is interpreted as a kernel $kern to be convolved with the $data.

If called in a list context, the filtered data and kernel ($dataf,$kern) are returned.


  $x = testdata($Npts, $freqs, $amps)

For example:

  $x = testdata(1000, [5,100], [1, .1] );

Generate a signal by summing sine functions of differing frequencies. The signal has $Npts elements. $freqs is an array of frequencies, and $amps an array of amplitudes for each frequency. The frequencies should be between 0 and 1, with 1 representing the nyquist frequency.


John Lapeyre, <jlapeyre at>


Copyright 2012 John Lapeyre.

This program is free software; you can redistribute it and/or modify it under the terms of either: the GNU General Public License as published by the Free Software Foundation; or the Artistic License.

See for more information.