[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Fwd: spectrum analyzer problem
From: |
Fabian Schwartau |
Subject: |
Re: Fwd: spectrum analyzer problem |
Date: |
Wed, 30 Jun 2021 16:03:58 +0200 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101 Thunderbird/78.11.0 |
OK, yes, I think this is a bug.
In
https://github.com/gnuradio/gnuradio/blob/master/gr-qtgui/lib/freq_sink_c_impl.cc#L553
The data is converted using the fft() function and averaged after that.
But the fft() function already converts the transformed data into the
logarithmic scale in
https://github.com/gnuradio/gnuradio/blob/master/gr-qtgui/lib/freq_sink_c_impl.cc#L359
Averaging after the conversion to dB will yield wrong values, as
log(a+b) != log(a) + log(b).
Fabian
Am 30.06.21 um 15:29 schrieb Ali G. Dezfuli:
>
>
> ---------- Forwarded message ---------
> From: *Ali G. Dezfuli* <ali69550@gmail.com <mailto:ali69550@gmail.com>>
> Date: Wed, Jun 30, 2021 at 4:56 PM
> Subject: Re: spectrum analyzer problem
> To: Fabian Schwartau <fabian@opencode.eu <mailto:fabian@opencode.eu>>
>
>
> thanks a lot, Fabian,
>
> I asked about the mapper itself and not pulse shaping (RRC). After some
> tests, I finally get to the problem of the spectrum analyzer (QT GUI
> Freq. Sink).
> I attached the grc photo to make my problem clearer.
> About the version, I really really like to be up-to-date but that's the
> matter of the OOT delay to move to the latest versions of GR and I
> really need them as well.😀
>
> On Wed, Jun 30, 2021 at 2:29 PM Fabian Schwartau <fabian@opencode.eu
> <mailto:fabian@opencode.eu>> wrote:
>
> Hi Ali,
> you should provide an example flow graph.
> And I think I heard several times on this list, that you should not use
> version 3.7 any more, as it is fairly old and not maintained any more.
> Nevertheless, I hacked together a small flowgraph, where I cannot
> reproduce your problem. I generate a QPSK constellation, the output RMS
> value is 1 (which makes the peak values slightly larger than 1!) and the
> 1000 point FFT with a rectangular window shows -30 dB. When you generate
> a constellation with 2 samples per symbol and make the symbol values to
> have a value of 1, you will probably get 3 dB less than expected, as the
> intermediate samples have less power than 1.
> Attached is the flowgraph and the results.
>
> Best regards,
> Fabian
>
> Am 30.06.21 um 11:19 schrieb Ali G. Dezfuli:
> > Hi all,
> >
> > I wonder if this is a bug or what:
> > I just make a QPSK constellation on the unit circle with quite random
> > data and
> > look at the spectrum by "QT GUI Frequency Sink" block in GRC.
> > With these parameters:
> > Â Â Â FFT size = 1000
> > Â Â Â Window type = None OR Rectangular
> > I should see a flat line as the spectrum with -30 dB (= -30 dBW) but
> > surprisingly it
> > shows -33 dB !!!
> > I use GNU Radio Companion 3.7.13.4 and ubuntu 16.04.
> > Would be grateful if you could help me in this matter.
> >
> > regards,
> > Ali
>