Re: [Discuss-gnuradio] SWIG related problem with new filter function

[John Andrews]

Here are all the related files.

C++ header

***************

* dsss_firdes.h *

***************

#ifndef _DSSS_FIRDES_H_
#define _DSSS_FIRDES_H_

#include <vector>
#include <cmath>

class dsss_firdes {
public:
  dsss_firdes();

static std::vector<float> raised_cosine (double gain, double sampling_freq, double symbol_rate,  double alpha,  int ntaps);
};

#endif

C++ Source file -

*****************

* dsss_firdes.cc  *

*****************

#include "dsss_firdes.h"
#include <stdexcept>

using std::vector;

// Raised Cosine Filter

vector<float>
dsss_firdes::raised_cosine (double gain, double sampling_freq, double symbol_rate, double alpha, int ntaps)
{
   ntaps |= 1; // ensure that ntaps is odd

   double spb = sampling_freq/symbol_rate; // samples per bit/symbol
   vector<float> taps(ntaps);
   double scale = 0;
   for(int i=0;i<ntaps;i++)
   {
       double x1,x2,rc1;
       double xindx = i - ntaps/2;
       x1 = M_PI * xindx/spb;
       x2 = 1-(4*alpha*alpha*(xindx/spb)*(xindx/spb));

      if(i == ntaps/2){
            taps[i] = 1;
            continue;
       }
      if(fabs(x2) < 0.000001){ // manage rounding errors. treat 0.000001 as zero
           x2 = 8*alpha*(xindx/spb)*(xindx/spb);
           rc1 = sin(x1)*sin(alpha*x1)/x2;
       }
       else{
       rc1 = (sin(x1)*cos(alpha*x1))/(x1*x2);
       }
      taps[i] = rc1;
      scale += taps[i];
      }

      for(int i=0;i<ntaps;i++){
           taps[i] = taps[i] * gain / scale;
        }
      return taps;
}

SWIG interface

** dsss.i **

%include "gnuradio.i"

%{
#include "dsss_firdes.h"
%}


//GR_SWIG_BLOCK_MAGIC(dsss,firdes);
class dsss_firdes
{
public:
static std::vector<float> raised_cosine(double gain, double sampling_freq, double symbol_rate, double alpha, int ntaps);
};

 ***************

*  Makefile.am   *

 ***************

include $(top_srcdir)/Makefile.common

# C/C++ headers get installed in ${prefix}/include/gnuradio
grinclude_HEADERS =  dsss_firdes.h

###################################
# SWIG Python interface and library

TOP_SWIG_IFILES = dsss.i

# Install so that they end up available as:
# import gnuradio.dsss
# This ends up at:
# ${prefix}/lib/python${python_version}/site-packages/gnuradio
dsss_pythondir_category = gnuradio


# additional sources for the SWIG-generated library
dsss_la_swig_sources =  dsss_firdes.cc

include $(top_srcdir)/Makefile.swig

# add some of the variables generated inside the Makefile.swig.gen
BUILT_SOURCES = $(swig_built_sources)

# Do not distribute the output of SWIG
no_dist_files = $(swig_built_sources)

On Sun, Jul 3, 2011 at 10:33 AM, Tom Rondeau <address@hidden> wrote:

On Sun, Jul 3, 2011 at 3:25 AM, John Andrews <address@hidden> wrote:

Hi Tom,

Are you doing this in your own top-level block (dsss from the looks of it) or under a current GNU Radio directory like gnuradio-core where the gr_firdes currently is?

Yes, I am doing this in my own top-level directory named 'dsss' (abbreviation for direct sequence spread spectrum)

You haven't mentioned adding anything to the Makefile so that it is built properly. If you are working under gnuradio-core/src/lib/general, you'll also want to add your .h and .i files to general.i so that it get's compiled.

I added the names to Makefile.am like I added the remaining names of files. The Makefile.am has been adapted from howto-write-a-block. The C++ part gets compiled without any problem but when importing in python it throws an undefined_variable_dsss_firdes_Z*** type of error, which according to my previous experience occurs with wrong SWIG interface problem. I am pretty sure I am having a SWIG related issue here.

Ok, it _sounds_ like you are doing everything correctly. In my experience with this stuff, it usually comes down to a small typo somewhere. You can post your files here (.i, .h, and Makefile) so others to maybe have a look to see if there's anything incorrect in them.

 

Also, just curious, what application are you interested in that requires a raised cosine filter? And couldn't you just create a root raised cosine filter and convolve the the taps against themselves to make the raised cosine taps?

I am developing a  BPSK based Direct sequence spread spectrum Tx/Rx and at the transmitter i wanted to use a raised cosine filter when interpolating from symbol to N samples before sending it to the USRP. I am not sure if this is the preferred method but I haven't found any documentation  related to DSSS that mentions pulse-shaping. 

Tom

There is definitely some work on pulse shaping of DSSS systems. I believe, though, that they all use root raised cosine filters. This is done at both the transmitter and receiver so that the received signal has gone through two root raised cosine filters to make the raised cosine filter, which is a Nyquist filter.

Tom