[Discuss-gnuradio] Theory - GSM traffic detection with a USRP board

[Carlo E. Prelz]

Good day everyone. There must be a wide variety of persons reading
this list. I hope there is some good soul there who happens to
concretely understand the wherefores of radio waves, and has the
ability to break the bread of this knowledge with those (like me) who
are more prone to logics than to electronics.

I write code for a living, and life has been so generous as to allow
me to deal with interesting stuff and apply them in the world of arts
(have a look at my simple site if you are interested - www.fluido.as).

I decided to engage in a project whose target is to detect the
existing cell-phone traffic in a specific surrounding and represent it
in some visually or acoustically interesting way. This is a personal
project, so I won't care too much if it gets to no concrete
result. But I invested in an expensive USRP board with appropriate
daughterboards (dbs-rx, range 500 MHz to 2.6GHz), so I would like to
at least give it a concrete try. The boards are connected to
log-periodic antennas provided by Ettus. 

Sadly, I did not notice in due time that the Gnuradio code base is
developed in C++/python. Had I seen this I would probably have not
started the project at all. I work in C/Ruby, a combination that is
decidedly at odds with the chosen languages of Gnuradio. 

I have had a very hard time extricating the logic for configuring the
USRP and setting the various parameters - frequency, gain, bandwidth,
decimation. But I am now reasonably certain that this part works OK
now in my code.

I am receiving a stream of data samples from the USRP. The
daughterboards return I and Q channels. I feed these to fftw, and
obtain the power spectrum of the required frequency band. So, for
example, with decimation=16 I receive a power spectrum covering 4
MHz. I confirm that this part is ok because I receive the same peaks
in the same areas with my software that I receive with the Gnuradio
tools. E.g: I have a peak at 832MHz (should be UHF 66 - here in
Holland most of TV broadcast is on cable, so there is little activity
on UHF) both with gnuradio-examples/python/usrp/usrp_fft.py and with
my application.

Now it is time to focus on the GSM bands. GSM900 should cover,
according to what I found on the net, 880MHz to 915MHz for the uplink,
in channels spaced 200kHz. What I found on this band was constant and
relatively strong signals spaced 4MHz - some stronger than others, the
strongest of them all at 896MHz and 900MHz. I could confirm that there
were carriers on these frequencies by listening on my hand-scanner.
But I then found out that those carriers would disappear when the USRP
was turned off. It has to be said that, to the scanner, the whole band
appears rather deserted (when the USRP is off). There is a carrier at
912MHz, but not much more. This last signal is indeed seen by the
USRP.

So, this is my question: I expected to find a lot of carriers active
at short bursts in the GSM900 band. This seems not to be the case,
even if there are lots of active cellphones around. Am I looking at
the wrong range? Or does the activity of cellphones manifest itself
somewhat differently, so that no traditional carriers can be detected?
And in this case, could you offer some pointers? I also had a look at
the 1.8GHz range (1710-1785 MHz), with similar results. These, I
believe, are the two bands that are used in this country for mobile
phone traffic.

I apologise in advance for my ignorance... I studied (long time ago)
as an architect.

Carlo

-- 
  *         Se la Strada e la sua Virtu' non fossero state messe da parte,
* K * Carlo E. Prelz - address@hidden             che bisogno ci sarebbe
  *               di parlare tanto di amore e di rettitudine? (Chuang-Tzu)