Re: [Discuss-gnuradio] Success (kinda)

[Rob Judd]

Alan,

A 60dB S/N is all you can expect from a tuner of this type. I don't
think that's the problem, that looks just like the spectrum I get on the
EA '92 Spectrum Analyser designed by Jim Rowe.

Rob


Alan Gray wrote:
> 
> >I'm assuming that you're tuned to a broadcast FM station, right?
> 
> Yes
> >
> >If you run adc3_fft and look at the spectrum, is there a nice
> juicy
> >spike where you expect your IF to be?
> 
> Yes, though the noise level surrounding the spike is larger
> than I would like (see attached screenshot radio_fft.png)
> 
> >
> >Have you experimented with the RF and IF gain settings for
> the tuner
> >module?  For lowest noise figure, you want to max out the RF gain
> >before ramping up the the IF gain.
> 
> The 4702 tuner only has one gain stage as it does a direct
> conversion from RF to the 36MHz IF. Essentially I change this
> gain to maximise the size of the peaks in the FFT while not
> amplifying the noise
> 
> >In fm_demod1 and/or hifi_fm did you you change the definition
> of the
> >IF frequency?
> 
> Yes, I've tried both 4MHz (where the spectrum will be
> reversed) and -4MHz (where it should not be reversed) to see
> if that made any difference, no luck though
> 
> >> Basically what I'm hearing sounds like a radio which is tuned
> >> to about 50KHz away from the actual carried frequency. I can
> >> make out beats and baselines but it sounds as if I'm missing
> >> the high end frequencies. I can make out when words occur, but
> >> again they're very distorted and it's hard to determine what
> >> is being said.
> >
> >Of the two wide band FM receivers we've got, hifi_fm sounds
> better
> >than fm_demod1.  In any event fm_demod1 doesn't suck
> completely (sort
> >of like a low cost hand held radio from the old days).  It's not
> >really processing a wide enough piece of the spectrum to grab
> it all
> >(related to the fact that there aren't nice integer
> decimation factors
> >that get us from the IF input sampling rate to the sound card
> output
> >rate).  hifi_fm interpolates, and does a couple of other
> things more
> >right, and as a result sounds quite good.
> 
> Ok, I guess the problem is somewhere in the hardware then as
> what I'm getting sounds nowhere near like even the most
> primitive radio
> 
> >
> >If anyone's looking for a not too hairy project, building a
> version of
> >hifi_fm.cc that decodes the L-R stereo info wouldn't be too hard.
> >
> >> Is this consistent with the results obtained by others or do
> >> you get a clearer output? If others do get a clearer signal,
> >> can anyone suggest a possible fix?
> >>
> >> Thanks
> >> Alan
> >
> >If you like I can send you some audio samples of FM I've
> >demodulated...
> >
> >Eric
> >
> 
> I actually grabbed the file with 5 seconds worth of FM data in
> it that was posted here a while ago and ran that through
> fm_demod1 to see what it sounded like. It was a whole lot
> better than what I'm getting so other audio sample are
> probably not necessary.
> 
> I've also tried to tune to frequencies a little to either side
> of my desired frequency to check if there is an issue with
> drift in the frequency of the output IF, and that doesn't seem
> to be the case. I'm starting to suspect that there may be an
> issue with the amount of noise introduced into the system.
> 
> I'm using a different ADC to the MC4020, an Adlink PCI9812,
> and I'm starting to think that might be having an impact on
> performace. The second attached image (no_connection_fft.png)
> shows the fft I'm getting with no signal input into the ADC.
> Seems to be a large amount of noise there unfortunately,
> though I'd like to compare it to a similar FFT using the MC4020.
> 
> If anyone has any suggestions on what it going wrong (and how
> to fix it) they would be most appreciated.
> 
> Alan
> 
>   ------------------------------------------------------------------------
>  [Image]  [Image]
> 
>   ------------------------------------------------------------------------
> _______________________________________________
> Discuss-gnuradio mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/discuss-gnuradio