[Octave-bug-tracker] [bug #66024] (signal) cl2bp crashes Octave 9.2.0 on Windows 10

[Dennis Cote]

Follow-up Comment #5, bug #66024 (group octave):

The Matlab equivalent function fircls (see
https://www.mathworks.com/help/signal/ref/fircls.html) does not require a
gridsize argument for the same algorithm. 

The authors of the original paper cited in the cl2bp documentation published
an improved algorithm in a later paper that is available at
https://www.researchgate.net/publication/3316722_A_Modified_Algorithm_for_Constrained_Least_Square_Design_of_Multiband_FIR_Filters_without_Specified_Transition_Bands.


This paper mentions a grid of frequency values used for the algorithm but
says: "When implementing the algorithm described herein, the frequency
response can be calculated over a grid of frequency values. However, it is
advised that the location of the extremal frequencies be refined by Newton’s
method; otherwise, a rather dense grid is sometimes required for convergence.
The use of Newton’s method is easily incorporated."

At the very least a default value of perhaps 2^8 or something similar should
be provided and documented.


    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?66024>

_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/

signature.asc
Description: PGP signature