Re: [Discuss-gnuradio] FFT-based feature extensions, fixes, etc.

[Eric Blossom]

On Fri, Oct 21, 2005 at 05:22:23PM -0400, Weber, Michael J. (US SSA) wrote:
> Has anyone done any work on creating some intelligent behaviors based on
> the FFT outputs? I'm thinking of things like autotuning to new strong
> signals, autoconfiguring demodulator bandwidths (wide or narrow FM, for
> example), drawing a vertical indicator line in the plot to indicate
> various things, etc. If nothing's been done, I'll plow ahead slowly...
> but no since recreating the wheel.

Please dive in.  
In truth, I think the right answer is to completely revisit the guts
of the plot widget we're currently using.  It would be most
appreciated.

> On a related note, is there a way to calm the auto-scaling behavior of
> the FFT plots?

Use the code in CVS, or wait a bit for 2.6.  It's doesn't autoscale
(It was driving us crazy too!).  With the new code, right clicking
gets you a menu that controls dB/div, ref level, averaging, etc.

> I've created a narrow-band version of the usrp_wfm demo,
> but the frequency increment is 5KHz instead of 100Khz, and the plot
> moves from left to right across the graph before the automatic scaling
> switches ranges. Sometimes the total range of X values shifts as well,
> from 4 to 5 MHz and back. I suppose what I'm looking for is more control
> of the scale/range of the plots, the grid lines, everything. What's the
> best way to proceed?

You probably want to look at the usrp_wfm_rcv.py code in CVS.  It's
quite a bit different from the code in 2.5.

The biggest contribution to the project would be to rewrite the
underlying plotting widget. I suggest something simple and
deterministic like a grid with fixed X & Y divisions.  Don't try to
autoscale, etc., just provide controls to scale X & Y.  FWIW, the
usrp_oscope manually manages the X axis, but autoscales Y.  It was
done by kludging around some of the underlying widget behavior.

> I'm working with a TVRX board at the moment, but I suspect that doesn't
> matter.

Shouldn't matter.

> Thanks for any hints and tips!
> 
> Best,
> Mike

Eric