FM Stereo decoder, frequency doubler, IIR issues

GNU Radio
1

Hi list,

I just got started with GNU Radio Companion for use with my FCDPP and
it’s
been great, already am decoding FM stereo and it sounds very good.
However
it probably isn’t the most optimal way to do it:

My filters are the inbuilt bandpass - they’re a bit hard on the CPU and
was
hoping to use the IIR filters as biquads instead but I can’t seem to get
them to work. Filter was designed with this applet (
http://www.earlevel.com/main/2010/12/20/biquad-calculator/ ) and I fed
the
A values into Feed-Forward of the IIR and B values to the Feed-Backward
and
it would not pass signal so I swapped A and B values and it apparently
just
passed the signal unaffected. Any ideas what I’m doing wrong?

What is the internal resolution for the IIRs? I am asking because some
of
my filter Q’s are in the 10-60 range and am concerned about blowing them
up.

Also, what would be the most efficient way (in GNU Radio) to get the
absolute value of a signal? The best I’ve come up with so far was a
Float->Complex feeding a Complex->Magnitude.

Thanks!

Chuck

2

On Wed, Apr 24, 2013 at 8:56 AM, Chuck R. [email protected]
wrote:

Hi list,

I just got started with GNU Radio Companion for use with my FCDPP and it’s
been great, already am decoding FM stereo and it sounds very good. However
it probably isn’t the most optimal way to do it:

There are FM receivers/demod blocks already that you can play and
experiment with. In 3.6, they are located in
gnuradio-core/src/python/gnuradio/blks2impl, look for wfm_rcv blocks.

Obviously, though, nothing stopping you from building and
experimenting on your own.

My filters are the inbuilt bandpass - they’re a bit hard on the CPU and was
hoping to use the IIR filters as biquads instead but I can’t seem to get
them to work. Filter was designed with this applet (
http://www.earlevel.com/main/2010/12/20/biquad-calculator/ ) and I fed the A
values into Feed-Forward of the IIR and B values to the Feed-Backward and it
would not pass signal so I swapped A and B values and it apparently just
passed the signal unaffected. Any ideas what I’m doing wrong?

I think this has already been addressed in another email thread, but
try negating the A vector. And try using the gr_filter_design tool.

What is the internal resolution for the IIRs? I am asking because some of my
filter Q’s are in the 10-60 range and am concerned about blowing them up.

The taps are passed in as double-precision floating point.

Also, what would be the most efficient way (in GNU Radio) to get the
absolute value of a signal? The best I’ve come up with so far was a
Float->Complex feeding a Complex->Magnitude.

Well, you could always write a block to do that yourself. There’s also
the blocks::transcendental block that you can use. Should be able to
pass it “fabsf” and “float” as the arguments.

Thanks!

Chuck

Tom