Re: octave and FIR plotting anomaly

[Mike Miller]

On Mon, May 19, 2014 at 22:58:35 +0300, Alfred Liin wrote:
> i have made a script to plot highpass and lowpass FIR filter frequency
> response (code is in the end of the mail). It used to give me the plot
> attached (madal_ja_korgpaas_koos4.jpeg)
>
> Now, i think there has been some update to octave and when copy-pasting the
> same code, i get another result (not as pretty graph - attached as
> Graph_problem.jpeg)
>
> Can you explain what is the problem? Which graph is correct?
> [...]
> octave:209> version
> ans = 3.8.1
>
> There are some unusual characters in my language that octave is no longer
> able to plot, but that is not the main problem. The main thing is that
> highpass filter has a different response in the new graph.
>
> Can you please check out the code:

Thanks for your very thorough question, but no need. This is a known
bug in version 3.8.1 that impacts the fir1 and fir2 functions. The
regression has been fixed [1],[2] and the fix will be released as a
part of the next Octave release (3.8.2 or higher).

If you want to fix this without waiting for the next release, you can
download the interp1 function from the working stable branch of Octave
[3] and add it to your runtime path locally to fix this problem. I've
added unit tests to the fir1 and fir2 functions to prevent this from
regressing in the future.

  [1] http://savannah.gnu.org/bugs/?41980
  [2] http://savannah.gnu.org/bugs/?41958
  [3] 
http://hg.savannah.gnu.org/hgweb/octave/raw-file/a6d13c62fccf/scripts/general/interp1.m

HTH,

-- 
mike