[Discuss-gnuradio] ssb demod

[mike revnell]

I got my usrp last month and have been learning my way around gnuradio 
and wxpython.

The attached tarball has a couple files which may be useful in examples. 
They implement ssb demod
in the frequency translating FIR.  It probably isn't the most efficient 
way to implement ssb but I think it is
interesting for a couple reasons.

It loads complex coefficients from a file. The coefficients were 
generated by the Scilab program
 which is also provided.

It implements a complex hilbert transform with symmetric I and Q 
branches. The more standard
 method is to compute the hilbert transform in one branch with a 
compensating delay in the other.

The phase processing is combined with low pass amplitude filtering in 
the same filter.

I'm not terribly familiar with the DSP literature but I've not come 
across exactly this before.

I haven't tried to quantify its performace in any formal way but it 
seems to work ok for listening in on
the 40M ham band. Tuning it across a cw carrier indicates that it does 
indeed reject most of the
undesired side band. There is some leakage at zero beat due to the 
finite transition band between positive
and negative frequencies.

The starting point for most of this was weaver_isb_am1_usrp3.py by cswiger.

Scilab is a free, very powerful, math package much like Matlab.

As background:

I'm in the correlator group here at NRAO in Socorro (at the VLA that 
starred
in "Contact"). I'm designing the digitizers and IF data transmission 
system for the EVLA project.

http://www.aoc.nrao.edu  Look for the EVLA project.

We are digitizing 4 IF channels per antenna at 1GHz bandwidth to 8 bits 
and delivering
120 Gbits/sec per antenna (up to 32 antennas) to the control building.

Eventually we'll be digitizing 8 IFs per antenna at 2GHz bandwith at 3 
bits resolution.