Re: [Discuss-gnuradio] gqrx branch osmosdr

[Alexandru Csete]

On Wed, Jun 6, 2012 at 12:29 AM, Patrick Strasser
<address@hidden> wrote:
> Alexandru Csete wrote on 2012-06-05 19:06:
>> On Tue, Jun 5, 2012 at 6:51 PM, Patrick Strasser
>>> In comparison to the rtlsdr fork I see you added FM-W. FM-N in
>>> comparison sounds more clipped, FM-W is clipped with a narrow filter,
>>> but noisy with a wide filter. With a filter that has no clipping, noise
>>> is clearly audible. I'm not sure if this is bound by the low dynamic
>>> sampling range (8 bit) in combination with big bandwith which means a
>>> lot more noise energy with wide filters. Anyway, gqrx gets greater every
>>> time!
>>
>> There is no audio filter yet (except de-emphasis) so you get pretty
>> much 48 kHz worth of noise, including stereo pilot tone and whatever
>> crap they include in a broadcast FM channel these days.
>
> It's AM and some special BPSK. Nice ensemble of different analogue and
> digital schemes, maybe we'll use it as an demonstration in one of our
> lectures here at University.
> Anyway, my ears have a builtin filter that cuts dramatically everything
> above some 18kHz, not sure if I can hear the pilot tone. If you are not
> doing dirty tricks every decimating filter should prevent aliasing, so
> no effect of everything that's higher than half the sampling rate. With
> 48 kHz you should not get even close to the higher energy parts of AM at
> arround 38kHz.
> I guess where noise comes in is before FM demodulation: the bigger the
> sampled frequency bandwidth, the more noise energy. With the limited
> dynamic range this decreases the signal (FM) to noise (background)
> ratio, which opposes good demodulation results.
> Per design and regulation a FM radio signal has mono 180 kHz bandwidth
> and stere 300 kHz, but I do not know of a station sending in mono here
> in Austria. Maybe a kind of matched filter would be best: you can expect
> that the wider fare a frequency is from center frequency, the less
> energy it has. Maybe thirds would work: < -150 kHz stop, up to -50 kHz
> transition, -50 to +50 kHz pass, +50 up to +150 transition, and then
> again stop. Frequency deviation is specified as 75 kHz.

Thanks for the info. I suspect reception quality will also depend on
the device, gain settings and how much interference there is.

Anyway, I have increased the sample rate of the WFM receiver to 240
ksps, done some tweaking of the resamplers (some would say I fixed a
major screw-up) and I think it now sounds quite all right. I can even
use the Funcube Dongle for WFM and get decent audio with minimal
distortions. Here is a short video showing hotswap between RTL device
and Funcube Dongle on the same WFM station:
http://www.youtube.com/watch?v=O9pLDLvOsPk

Alex