Re: [Discuss-gnuradio] Harmonics of the sampling freqeuncy

[John Orlando]

On Fri, Aug 20, 2010 at 9:03 AM, devin kelly <address@hidden> wrote:
> Hello,
>
> I'm doing a spectrum sensing project and I'm testing both the USRP 1
> and USRP 2.  So far, I've modified usrp_scectrum_sense.py to do use
> the USRP2 with the WBX
>
> Right now I'm doing this is what I'm doing with what gr_bin_statistics
> gives me (as discussed in a previous post)
>
>         signalPower=0
>         const = -20*math.log10(tb.fft_size)-10*math.log(tb.power/tb.fft_size)
>         for bin in m.data:
>                 signalPower += 20*math.log10(bin/tb.fft_size) + const
>
>
> I then save each of the signalPower variables to a file and plot them
> using matlab.  The attached is what I get when I plot, this is the
> results of one sweep with the USRP2 and WBX.  I have several questions
> about my results.
>
> 1) There are harmonics at multiples of the sampling frequency (100
> MHz) which I have pointed arrows at.  Has anyone encountered this
> before?  Are there any solutions?

Yes, we've seen the harmonics at multiples of the USRP2 clock rate
(which is 100 MHz).  I don't know of any solution to this issue but
would be interested if anyone else has conquered it.  We spent a good
bit of time trying different options to remove these spurs as we were
developing our BURX daughterboard, but ultimately they seem to be
radiating from the USRP2.  From our experiments, there is also a small
component (~10 dB) that is coming in through GND as well, but the
dominant source is radiated.  Shielding on the WBX daughterboard may
help (I believe I've seen copper tracks around the perimeter of both
the Tx and Rx sections on the WBX that look like you could solder down
a shield of some sort...assuming these stitch down to a GND layer, you
may be able to provide an RF can around the RF portions of the
circuits, and this could help reduce these received spurs).

>
> 2) This sweep was done with a constant gain (31) on the WBX.  Should I
> be changing the gain as a function of the frequency I'm scanning?  I
> ask because after 1.5 GHz power seems to decrease quickly.

Yes, to maintain the optimal performance of the receiver, the gain
should be changed based on the frequency of operation.  For optimal
noise figure, Ettus Research posted a series of plots of gain
configuration vs. frequency that showed the optimal gain setting at
each frequency of operation to also maximize IIP3.  See:

http://code.ettus.com/redmine/ettus/attachments/46/imd_and_nf_vs_gain.pdf

for details at a handful of frequencies.

>
> 3) In this sweep I had a terminated all ports with 50 ohm terminators,
> so I would expect my sweep to be fairly flat.  However, there peaks
> and troughs.  Does anyone know how I could get a flatter response?
>

You should be able to get a "flatter" gain throughout by more
carefully selecting your gain settings.

-- 
John Orlando
CEO/System Architect
Epiq Solutions
www.epiq-solutions.com