Hi!

I fully agree with "for modulated signals, power estimates get more
complicated"; I have the uncontrollable urge to be a smartass, however,
about this:


> A sine wave only has one bin.

Nope. That's true only for oscillations that fit /exactly/ in the number
of samples observed by the DFT you're doing. For example:
Assume you do a 1024-FFT¹.
When your sine has a period of let's say 64, everything is fine. You get
an DFT that's all zeros but for one bin, because the oscillation that
repeats 16 times within one DFT correlates perfectly with that (so you
get one peak for the negative- and one peak for the positive-frequency
complex sinusoid).
But suppose you do a sine of periodicity 65! Now, there's no single bin
that correlates "perfectly", because 65 is not a factor of 1024. But:
Parseval's Theorem states that the energy can't disappear. In fact, that
energy ends up in neighboring bins.

Which, by the way, is the reason why it's dangerous to trust a DFT-based
spectrum estimate, unless you know that the signal bandwidth is
significantly larger than the bin spacing. You can only achieve that
through increasing the DFT (FFT) length, and that in turn increases the
observation time.

Best regards,
Marcus

¹NB: DFTs /FFTs do /not/ have to be powers of two in length. There's
plenty of FFT implementations for length that can be factorized to
"small" prime factors, and you can do any (non-FFT) DFT length, albeit
at higher computational cost. So, a 1023-DFT is perfectly possible to have.

On 26.06.2017 11:02, li...@lazygranch.com wrote:
> When a signal is modulated, you need to use a FFT technique ‎to get
> the power. My experience is all based on outmoded analog modem design,
> but the idea is the same. The hardware FFT based spectrum analyzer
> sums up the power in all the bins and gives you a number.  
>
> A sine wave only has one bin. So a modulated signal will always have
> lower peaks than a sine wave for the same power.
>
> I have a HackRF and appreciate  this work.
>
> *From: *Ralph A. Schmid, dk5ras
> *Sent: *Sunday, June 25, 2017 11:44 PM
> *To: *'Murat Aksu'; Discuss-gnuradio@gnu.org
> *Subject: *Re: [Discuss-gnuradio] dB or dBm
>
>
> Well, dBm is an absolute power, based on 0dBm = 1mW. The dB figures of
> your receiver are only relative values, they have no meaning. However
> those get interesting when something changes. You just need to
> understand that the input of -20dBm power has nothing to do with the
> resulting reading of -58dB. When you reduce the input power for
> another 10 dB (no matter if you turn it down or if you insert an
> additional 10dB attenuation, makes no difference), then your reading
> should change from -58dB to -68dB.
>
>  
>
> Then you have found a part of the linear range, where a direct
> relationship exists between the input power and the power reading.
>
>  
>
> AGCs may influence this, also nobody knows how accurate the peak power
> measuring works. As others already had mentioned, it could be useful
> to play with unmodulated carriers, just to get a feeling for this stuff.
>
>
> Ralph.
>
>  
>
> *From:*Murat Aksu [mailto:muratc...@hotmail.com]
> *Sent:* Thursday, June 22, 2017 4:58 PM
> *To:* Ralph A. Schmid, dk5ras
> *Subject:* Re: [Discuss-gnuradio] dB or dBm
>
>  
>
> Dear Ralph,
>
>  
>
> Thank you so much for your support. I really do not understand these
> dB values. When I inject 802.11g signal with -20 dBm power level and
> 20 dB attenuator which means -40 dBm going in the HackRF One, after
> running gr-scan command line or QSpectrumAnalyzer GUI, I am observing
> peak power values around -58 dB.
>
>  
>
> As you suggest, if I use 30 dB attenuator instead of 20 dB attenuator,
> I will still get some dB values and really don't know how they are
> related to power levels.
>
>  
>
> I would appreciate your guidance about this confusion.
>
>  
>
> Best,
>
> Murat
>
>  
>
> ------------------------------------------------------------------------
>
> *From:*Ralph A. Schmid, dk5ras <ra...@schmid.xxx
> <mailto:ra...@schmid.xxx>>
> *Sent:* Thursday, June 22, 2017 7:38 AM
> *To:* 'GNUBeginner'; Discuss-gnuradio@gnu.org
> <mailto:Discuss-gnuradio@gnu.org>
> *Subject:* RE: [Discuss-gnuradio] dB or dBm
>
>  
>
> But still the dynamic range usually ends already quite below the maximum
> allowed level what is more a kind of hardware protection rule.
>
> Add some more attenuation and see if at least a 5dB change on your
> transmitter creates a 5dB change on your SDR. Then you are in the linear
> range and can start trying to calibrate the thingy for your frequency. And
> turn off AGC if this option is available.
>
> Ralph.
>
> > -----Original Message-----
> > From: Discuss-gnuradio [mailto:discuss-gnuradio-
> > bounces+ralph=schmid....@gnu.org
> <mailto:bounces+ralph=schmid....@gnu.org>] On Behalf Of GNUBeginner
> > Sent: Wednesday, June 21, 2017 10:48 PM
> > To: Discuss-gnuradio@gnu.org <mailto:Discuss-gnuradio@gnu.org>
> > Subject: Re: [Discuss-gnuradio] dB or dBm
> >
> > I am already aware of what the maximum allowable receiver power which is
> -5
> > dBm. That is why I am starting from 0 dBm with 20 dB attenuator before
> > injecting it to the HackRF One.
> > 
> >
> >
> > --
> > View this message in context: http://gnuradio.4.n7.nabble.com/dB-or-dBm-
> > tp64323p64335.html
> > Sent from the GnuRadio mailing list archive at Nabble.com.
> >
> > _______________________________________________
> > Discuss-gnuradio mailing list
> > Discuss-gnuradio@gnu.org <mailto:Discuss-gnuradio@gnu.org>
> > https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
>
>
>
>
>
>
> _______________________________________________
> Discuss-gnuradio mailing list
> Discuss-gnuradio@gnu.org
> https://lists.gnu.org/mailman/listinfo/discuss-gnuradio

_______________________________________________
Discuss-gnuradio mailing list
Discuss-gnuradio@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio

Reply via email to