Re: [Discuss-gnuradio] Why FM doesn't sound good

[Eric Blossom]

On Fri, Dec 21, 2001 at 12:30:53AM -0800, Matthew Ettus wrote:
>       I believe that the reason FM radio doesn't sound as good as it should
> is that the demod bandwidth is too low.  In the example code, the 20MHz
> sample rate is decimated by 125, to 160ksamples/sec (complex), for less
> than 160kHz BW.  To get good fidelity you really need 200 to 250 kHz, or
> the signal will be distorted ("soft" clipping).

Thanks for looking at this.

>       Unfortunately, since we are constrained by soundcard output rates, the
> choice of decimation rate is not free.

I think with AC97 codecs the choice is actually free.  I believe that
they will resample up to 48 kHz on the fly.  I haven't tried it; just
noted it in some spec I read a while ago.

> I came up with the following decimation plans:
> 
> 1)    Dec by 75, demod, interpolate by 3, decimate by 25, for a 32kHz
> output rate
> 
> 2)    Decimate by 50, demod, interpolate by 3, decimate by 25, for a 48 kHz
> output rate.
> 
>       The second one will let a little too much adjacent channel noise in,
> and have a higher computational load.  However, 48 kHz is a better
> sampling freqency for high quality FM -- since FM doesn't cut off until
> after 16 kHz it needs more than 32ks/s.

Is the adjacent channel noise because the default windowed FIR doesn't
have steep enough skirts?  As I recall, the Q of the windowed filter
(normalized narrowness of pass band) is linear with number of taps.
I.e., twice as many taps ==> twice as narrow of pass band.  We ought
to be able to get a narrower pass band (if reqd) by increasing the
number of taps (with concommitant increase in mips).

>       Tomorrow I'm going to try the 1st (since my computer is too slow for
> the second).  I need to check whether floating point sample rates are
> allowed in pSpectra, since the intermediate sample rates are
> non-integral.

Floating point sample rates are OK.  Let me know if you want me to try
anything on my set up.

> Matt
> 
> 
> P.S.  I updated the VrAudioSource and VrAudioSink to (optionally) take a
> char * with the name of the device.  This is useful for those like me
> with multiple soundcards, USB sound dongles, etc.  It's backwards
> compatible, BTW.

Thank you.

Eric