Re: [Discuss-gnuradio] improper WBX transmission of tone in center of spectrum

[Matt Ettus]

On 08/04/2010 09:17 PM, George Nychis wrote:
> Thanks for the help, Matt.  I never actually knew what DC offset was
> until now ;)  Learning as I go!
>
>  From what I understand, the USRP2 scales between -1.0 and 1.0.  I am
> trying to boost my transmit power well over the power of the DC offset.
> I think that if I increase the gain, the DC offset is also boosted.  So
> I am trying to boost my signal over it (for this 75dB separation).  I am
> scaling my samples to the -1 and 1 range, but I am not seeing that much
> separation after scaling:
> http://cyprus.cmcl.cs.cmu.edu/~gnychis/boost.jpg
>
> That's about a 5dB separation after scaling.  I am not adding any
> attenuation, and this is conducted.  I am also configuring the gain to
> 0, so as to not boost the DC offset either.
>
> Thanks for the help, I appreciate it.
>
> - George

George,

There is key difference between tones and continuum power.  The DC 
offset spike in your display shows up at about -58 dBm in the display. 
Since it is effectively a pure tone and all at one frequency, this is 
the correct power -- there really is -58dBm of power in that tone.

The peak shown for amplitude of the OFDM signal is about -46 dBm.  That 
does NOT mean that the OFDM signal power is -46 dBm, since the signal is 
spread over a wider bandwidth.  From the display it looks like it is 
spread over about 300 kHz.  Since your resolution bandwidth setting 
(RBW) is 10 kHz, this means that EACH 10kHz has about -46 dBm in it.  To 
get the total power you need to integrate over the whole bandwidth. 
This is easy with OFDM since it is rectangular.  Since 300 kHz is 30 
times 10kHz, there is about 30 times as much power in total.  30 is 
about 15 dB, so your desired signal is about -31dBm.

You can visually demonstrate this by changing the RBW on your analyzer. 
 If you increase it by a factor of 10, then the OFDM signal will rise 
on the display about 10 dB, while the DC tone will stay at the same 
level.  If you decrease the RBW by a factor of 10, the OFDM will drop 
10dB on the display, but the tone will stay the same.

One easy way to find the total power in your signal is to set the RBW to 
wider than your whole signal, so that everything fits inside it.  Then 
you can read total power by looking at the peak on the display.  This 
only works if your signal is narrower than the max RBW, typically 3 MHz.

Incidentally, this works the same with an FFT display -- the RBW is 
roughly the width of a bin for typical window functions.

A -31dBm signal would mean that the DC offset is about 27dB (not 5 dB as 
it looks on the display) weaker than the DC offset (aka LO leakage) of 
-58 dBm.  You can likely increase your digital amplitude without 
increasing the LO level, to get better than this 27dBc number.  To 
further reduce the LO leakage, you need to compensate for the natural DC 
offset.  This can be done on the USRP1 or USRP2 by changing the TX DC 
Offset parameters.


Matt