USRP2 Sample 2.4 GHz Channel

GNU Radio
1

Hello,

I’m trying to sample the 802.11b wireless channels but the USRP2. I’m
currently using revision 10689 of Gnuradio.
I’ve added code to bbn_80211b_rx.py to connect the gr_probe_signal_f()
to
the top block.
To sample the channel, I use gr.probe_signal_f().
Here’s the code to connect the block:

self.connect(self.u, self.conv_c2f, self.probe2)

Then I use the gr.probe_signal_f().level() to retrieve the signal in a
loop.

data_file = open('data.txt', 'w')
T1 = time.asctime()

while cs_samples < 1000000:
    CST = self.tb.probe_channel()
    data_file.write(str(CST) + '\n')
    cs_samples += 1

    data_file.close()
T2 = time.asctime()

def probe_channel(self):
“”"
“”"

return self.probe2.level()

I was wondering if this is the fastest way to sample the channel from
python. When I compute how long this process takes, it appears that I
can
get a sample every 8-9 microseconds, which seems a lot slower than what
the
USRP2 is built to do. Am I sampling the channel in an incorrect way?
I’m running this on a machine with a Intel Quad Q9650 @ 3GHz.

Thanks,
Miklos Christine

2

When you say sample the channel - are you trying to look at the I&Q
samples coming right out of the USRP2? In which case, the easiest way
to start would be to use the usrp2_rx_cfile.py script, then you can
load the file into Matlab/Octave/etc. to take a look at. If you want
to record samples coming out of one of a block in the flowgraph (e.g.
in the 802.11b scripts) - you can modify the flowgraph to connect up a
file_sink at each point you want to record data from.
Doug

On Tue, Sep 15, 2009 at 9:41 PM, Miklos Christine
[email protected] wrote:

    data_file.close()

get a sample every 8-9 microseconds, which seems a lot slower than what the

–
Doug G.
[email protected]

3

Yes, that is exactly what I’m looking for. However, when running
usrp2_rx_cfile.py with decimation = 4, I get a ‘S’ printing to stdout.
Is
anyone else able to run usrp2_rx_cfile.py and not get that overrun
message?
What could be the reason for ‘S’? Is it a limitation on how fast we can
write the samples to the hard disk?
I want to keep the maximum number of samples possible.

Thanks,
Miklos Christine

On Wed, Sep 16, 2009 at 6:52 AM, Douglas G. <

4

Depending on how long you want to sample (and how much system RAM
you’ve got) you can also do something like using tmpfs (i.e. a RAM
disk). I’ve often resorted to this when doing collections lasting a
minute or less. At 25MS/s * 32bit/S (I&Q@16bits) * 60s = 5.58GB
according to my calculator - so that may or may not be a viable option
for you.
Doug

On Thu, Sep 17, 2009 at 1:00 AM, Matt E. [email protected] wrote:

At a decimation rate of 4 you are generating 25 MS/s. If those samples are
complex shorts, that works out to 100 MB/s. If those samples are complex
floats, that is 200 MB/s. In the former case, you really need a RAID array
of at least two fast drives to keep up. In the latter, you’re going to need
to do something more exotic.

Matt

–
Doug G.
[email protected]

5

Miklos Christine wrote:

Yes, that is exactly what I’m looking for. However, when running
usrp2_rx_cfile.py with decimation = 4, I get a ‘S’ printing to stdout.
Is anyone else able to run usrp2_rx_cfile.py and not get that overrun
message?
What could be the reason for ‘S’? Is it a limitation on how fast we can
write the samples to the hard disk?
I want to keep the maximum number of samples possible.

At a decimation rate of 4 you are generating 25 MS/s. If those samples
are complex shorts, that works out to 100 MB/s. If those samples are
complex floats, that is 200 MB/s. In the former case, you really need a
RAID array of at least two fast drives to keep up. In the latter,
you’re going to need to do something more exotic.

Matt