Re: [Discuss-gnuradio] halfband in CVS

[Eric Blossom]

On Tue, Oct 25, 2005 at 01:17:37PM -0400, Lamar Owen wrote:
> Attached are some captures of plots made with a slightly customized
> usrp_fft.py (and older version, too; not sure if that impacts the curve
> shapes).  The customization is changing the average_alpha to 0.001 as well
> as the number of fft points to 1024, and the scales.
> 
> The curve shapes are quite interesting.  Perhaps tweaking of some
> coefficients somewhere in the halfband verilog would make it flatter
> across more decimation ratios?  The current best flatness is at a decim of
> 32, as is clearly visible in the plots.

32 looks great.  Your 64 and 128 look like what the 10 looked like for
me.  Let me rerun the experiment on my end.

Eric