Artefacts in usrp_fft.py

GNU Radio
1

Hi all,

We implemented a spectrum senser based on GnuRadio for our cognitive
radio
research. In order to calibrate the senser we check the performance of
the
RFX2400 and DBS RX d’board with the usrp_fft.py script. We use the norma
FFT
and the waterfall plot of the script. Feeding the usrp with generated
signals
by a high performance signal generator we could observe the appearance
of
some strong artefacts in the spectrum. To find out the origin of this
artefacts we put a calibrated 50 ohm termination on the sma antenna
connector
in order to assure we only measuring the noise floor of the usrp. Now we
could observe several artefact in almost all frequencies between 800 and
2500
MHz. Sometimes the artefacts are sharp peaks some times wider peaks and
the
total number of peaks varies also from center frequency to center
frequency.
Normally we use a decimation rate of 8 to sense a chunk of 8 MHz and we
try
several gains of the d’board. If someone know how the origin of these
artefacts and/or how to avoid getting them in the sensed spectrum please
answer. Every advice will be appreciated. Thank you all in advance

Luis Simoes

2

Hi Luis,

Google “receiver spurs” or “rf spurs” or “a/d spurs”. Those kinds of
things
may be the culprit. One test I often do if I suspect a spur is note the
frequency of the spur, tune the receiver by say 1khz, and then look at
the
frequency of the spur again. If it doesn’t move or it moves in the
opposite
direction or it moves an amount other than 1khz, than it’s an internal
artifact and not external. Can you connect a spectrum analyzer in front
of
the A/D and see if the spurs are there? You might also consider
histograming
your noise-only input. The histogram should be Gaussian shaped and
zero-mean. Any sort of A/D problem could show up in the histogram. For
example, spikes in the histogram may indicate bits that are “stuck”.
Also,
A/D’s tend to have quantization effects that are periodic and so show up
as
spurs. Adding dithering noise helps to break up the periodicity and
reduce
the spurs. I’d be interested in what you find.

Mike