Re: [Discuss-gnuradio] Computing Power Spectra

[John Orlando]

On Tue, Mar 15, 2011 at 4:48 PM, Scott Johnston
<address@hidden> wrote:
> Hello,
>
> I wrote a simple grc graph to compute power spectra of the output of the
> usrp. My test setup is a signal generator outputting a tone at the same
> frequency that I set as the usrp center frequency, directly connected to the
> usrp. But when I examine the output in Matlab, the first vector is exactly
> what I expect, a peak at the center and then the noise floor. The following
> vectors show the peak at lower and lower power levels until, after 10 or so,
> it disappears into the noise. My code is attached. Anybody know what could
> be causing this?

My guess is that you are seeing a DC offset correction loop closing.
This would occur if the RF tone of interest is downconverted to DC
(which sounds like what you described).  I don't recall how DC offset
correction is actually implemented in USRP land, but it may very well
take a few tens/hundreds of samples before it is completely corrected,
which lines up with what you're seeing.

Try changing your sig gen to output a tone at, say 100 Khz from your
RF tuning frequency, and it should not attenuate over time.

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