Re: [Discuss-gnuradio] Bug in freq_xlating_fir_filter_XXX

[Achilleas Anastasopoulos]

I will submit the patch.

regarding the sign change in frequency, I didn't mean to change the convention:

the sign change IS REQUIRED in order to KEEP the convention because now
the taps are not reversed...

Achilleas

On Wed, Oct 9, 2013 at 9:20 AM, Tom Rondeau <address@hidden> wrote:

On Tue, Oct 8, 2013 at 9:39 PM, Achilleas Anastasopoulos
<address@hidden> wrote:
>
> I was playing around with
>
> fir_filter_XXX
>
> and
>
> freq_xlating_fir_filter_XXX
>
> and noticed that the two do not give the same output
> for the same input (and center_freq=0 in the xlating filter).
>
> Looking at the implementation of the latter
> it is obvious why: the taps are reversed in the line:
>
> std::reverse(ctaps.begin(), ctaps.end());
>
> For consistency the taps should not be reversed (as in all other filters)
> We also need to set

Yes, please submit a patch for this. The taps are reversed inside the
fir filters, so this is redundant and confusing. Most people probably
use symmetric filter taps, which is why it has not been found.

> float fwT0 = 2 * M_PI * d_center_freq / d_sampling_freq;
>
> (instead of the minus sign in the code).
>
> unless there is an objection, I will go ahead and push a correction,
> Achilleas

Don't change the sign of the frequency. I know this is controversial,
but from my experience with users, more people find the current way
easier to understand. We're telling the filter what the center
frequency is, which means that we will take a signal at Fc and
downshift it to DC. To me, if we're on carrier Fc and we specify -Fc
as the "Center Frequency", that's more confusing.

Tom