Re: [Discuss-gnuradio] GMSK channel filter

[Matt Ettus]

Ben Olsen wrote:
> Hello,
>
> I have a bit of a confusion concerning the decimating Channel filter
> in the stock receive_path.py script that I'm using with the GMSK rx
> and tx python script:
>
> The filter coefficients for the chan_filt filter are determined by the
> gr.firdes.low_pass() function, which takes, among other arguments, the
> midpoint and the width of the transition band.  In the gr_firdes.h
> file it is specified that the units for these arguments is Hz, but in
> the code it says that the midpoint of the trans. band is "1.0".  The
> script correctly receives data on a much larger band than 1 Hz, so
> what am I missing?  Is the value somewhere being changed to designate
> a different value at the filter's level?
In many instances we use a normalized frequency, since these are sampled
systems.  A normalized frequency of 1.0 is equal to the sampling rate. 
You can think of the units as cycles per sample.  The first nyquist zone
would be from 0 to 0.5 if these are real samples, or -0.5 to 0.5 if
these are complex samples.  You

In some other instances we may use radians per sample, which would be
2PI times the normalized frequency.

Matt