Re: [Discuss-gnuradio] FFT plot unit

[Marcus D. Leech]

On 08/03/2015 09:36 AM, Sylvain Munaut wrote:
> Hi,
>
> > I was able to gather results, and I am really confused with it. I generated
> > a -30 dB signal based on the fft plot shown and transmitted it using a usrp.
> > My spectrum analyzer received a signal at -50 dBm (-80 dB) and my receiver
> > which also uses a usrp received the signal and plotted it at -30 dB. My
> > question is, what is the unit of the fft plot? Is it dB or dBm?
> They're dBFS  or dB Full Scale ...
>
> So they're just relative to the "full scale" range defined in GR as -1.0 ... 1.0
>
> USRP are not calibrated instruments, so you can't map this to dBm or
> any absolute power measurement. Only relative measurements are valid.
>
>
>
> Cheers,
>
>     Sylvain
>
> _______________________________________________
In order for Gnu Radio to display calibrated power units, it would need 
to have a very vigorously-defined interface to each and every piece of 
hardware so
  that power levels displayed in the FFTs are in calibrated convenient 
power units, like dBm.   This in turn, would require that every piece of 
hardware that
  connects to Gnu Radio be vigorously calibrated over their entire 
operating parameter space, including sample-rate, tuning frequency, and 
gain setting.
  This isn't, as you might imagine, practical.

So, what Gnu Radio receives are digitized voltage samples that are 
mostly-linearly-proportional to the voltage received at the antenna 
terminals
  of the device.  These are in turn, for purposes of convenience and 
generality, converted into a floating-point number in the range {-1.0,+1.0}
  within a flow-graph.  But without calibration on the part of the 
end-user, they are "unitless", and you have to determine the proportionality
  in the context of your own application.