Re: Freqz behavior

[Rik]

On 01/16/2015 01:57 PM, Doug Stewart wrote:

On Fri, Jan 16, 2015 at 12:02 PM, Rik <address@hidden> wrote:

On 01/16/2015 05:31 AM, Doug Stewart wrote:
> On Thu, Jan 15, 2015 at 8:42 PM, Rik <address@hidden> wrote:
>
>     On 01/15/2015 05:35 PM, Doug Stewart wrote:
>>     On Thu, Jan 15, 2015 at 4:44 PM, Rik <address@hidden> wrote:
>>
>>         1/15/15
>>
>>         Could someone with access to the Signal Toolbox run the following:
>>
>>     I think Rik means some one with Matlab Signal Toolbox
>>
>>
>>         format long
>>         [a,b] = freqz (1, [1, -0.5, 0.8], 10)
>>
>>         Trying to resolve this bug https://savannah.gnu.org/bugs/?43895
where
>>         Octave is apparently off by 1 point.
>>
>>         --Rik
>>
>>
>     Thanks,  I've already been helped out by Tallis Huther da Costa.
>
>     --Rik
>
> How does Matlab handle this?

They do just what we thought.  When you request the output of freqz it
produces a vector with N elements.  When you don't request an output, and
want to have the filter response immediately plotted, they produce a vector
of N+1 elements.  I have a patch, but it only works if the "whole" argument
is not specified.  I'll send that to you and maybe you can see how to tweak
things.

[h,w] = freqz (b,a, N);     # h and w are N-point vectors
freqz (b, a, N);  # h and w are N+1-point vectors as determined by get (h,
'xdata') and get (h, 'ydata') on the created plot.

--Rik

What handle did you use to get the x and y size??

I tried

figure(3)

freqz1 (1, [1, -0.5, 0.8], 10 , "whole")

h = gca ()

get (h)

get (h, 'ydata')

And i get:

error: get: unknown axes property ydata

I have it working for "whole" but I want to verify it against Matlab before I finish it up.

freqz produces two plots within a single figure.  I grabbed the first one with

freqz (1, [1, -0.5, 0.8], 10)
subplot (2,1,1)
h = get (gca, 'children')
get (h, 'xdata')
get (h, 'ydata')

--Rik

p.s. At least for the 'half' argument Octave did a better job than Matlab and calculated the final point correctly.  Matlab gave an answer of Inf for the normalized frequency of 1.0 and the filter in the code.