Re: [Discuss-gnuradio] FFSK/MSK Demodulation in Gnuradio

[Michael Kramer]

Am 02.12.2015 um 23:53 schrieb Michael Ossmann:
> On Wed, Dec 02, 2015 at 02:41:29PM +0100, Michael Kramer wrote:
> > I've tried your suggestion but again I'm a little bit confused by the
> > results. The attached Wave2 is produced if I decimate by 100 in my Low
> > Pass Filter right before the Hilbert -> Quadrature Demod. If I don't
> > before, but after the second Quadrature Demod I get Wave3. So I guess
> > I'm doing right by decimating within the Low Pass Filter.
> Agreed.  Wave2 is very close to the result I was hoping for.  Try
> increasing the number of taps in the Hilbert block by 2x to 4x.
>
> > I can even more or less see my data inside Wave2 (even though it seems
> > to be inverted. it looks like its 00000111111111001110110 instead of
> > 111111000000000110001001), but I'm not really able to center it since
> > the upper spikes are crossing zero sometimes.
> You might try using DC Blocker for this.  You'll probably have to
> increase the default length by about 3x.
>
> > Also I wanted to ask if my calculation is correct: I have sampled with
> > 2MHz. The bitrate (should be) 2400. So if I want 4 or 5 samples per
> > symbol for the Clock recovery I need to resample to 4*2400=9600 or
> > 5*2400=12000.
> Yes.  For best clock recovery results, I usually aim for 2 to 4 samples
> per symbol.  If you decimate by 400 instead of 100, that should give you
> 5 ksps or 2.0833 samples per symbol.  (And, actually, if you do that,
> you may not need to set the parameters for Hilbert and DC Blocker so
> high.)
>
> Mike
I've played around a little bit with the values again and I've found
something interesting. Even though the specification tells me it has a
bitrate of 2400 I get wave5 after demodulating with that. But if I use
4800 I get wave4 which looks much better. So since the signal isn't FFSK
or MSK according to the specification I'm not sure the bitrate is corect
either. But even with the good looking wave4 I can't seem to use the
Clock Recovery right. I've used a decimation of 102 which should bring
me to 19231 samples, which roughly equals either 4 samples at 4800 or 8
at 2400. I also tried to decimate by 204 which brings me down to 9803
samples which roughly equals 2 samples at 4800 or 4 at 2400. But with
neither of these settings the clock recovery seems to work, even though
my wave now looks pretty correct so I'm kinda lost at what to do.

Greetings,
Michael


---
Diese E-Mail wurde von Avast Antivirus-Software auf Viren geprüft.
https://www.avast.com/antivirus

wave4.PNG
Description: PNG image

Wave5.PNG
Description: PNG image