Re: [Discuss-gnuradio] FIR filter in fsk_tx.py

[Matt Ettus]

Sachi wrote:
> Let me make my question more clear.
> If I understand the code correctly, after
> "bytes_to_syms", the signal should be a float sequence
> of +1 or -1.  I am very confused about what the signal
> will look like after "interp".  I think it's no longer
> +-1 sequence any more.  So does it have any special
> function here?

Interpolation by N inserts N-1 zeros between each incoming sample, and
then filters the sequence.  The resulting signal will now contain values
other than -1,0, and 1.

> For the FIR interpolation filter, why we choose
> 1.2*data_rate/2 as the cutoff frequency and
> 0.4*data_rate/2 as the transition band?

The 1.2 and 0.4 are somewhat arbitrary.  What you need is to have the
cutoff frequency be above the first image in the interpolation and below
the second.  The transition band should also end before the second image.


> Second, why k = 2 * math.pi * max_deviation / self.fs?

This is how we set the deviation in radians per sample.  max_deviation
is a frequency in cycles per second.  Multiplying by 2*PI gives us
radians per second.  Dividing by the sample rate gives us radians per
sample.

>>I am a little confused. Why do you use sw_interp as
>>the gain? 

The interpolation ratio is used as the filter gain.  The reason for this
is that performing an interpolation by N will reduce the amplitude in
the signal by a factor of N.  By using N as the gain, this effect is
canceled out.

Matt