This is a work-in-progress implementation of a m-sequence based channel 
sounder for GNU Radio and the USRP.

In typical use, the user would run the sounder as a transmitter on one
USRP, and a receiver on another at a different location.  The receiver
will determine the impulse response of the RF channel in between.

The sounder uses a custom FPGA bitstream that is able to generate and
receive a sounder waveform across a full 32 MHz wide swath of RF spectrum;
the waveform generation and impulse response processing occur in logic in
the USRP FPGA and not in the host PC.  This avoids the USB throughput 
bottleneck entirely.  Unfortunately, there is still roll-off in the AD9862
digital up-converter interpolation filter that impacts the outer 20% of
bandwidth, but this can be compensated for by measuring and subtracting
out this response during calibration.

The sounder is based on sending a maximal-length PN code modulated as BPSK
with the supplied center frequency, with a chip-rate of 32 MHz. The
receiver correlates the received signal across all phases of the PN code 
and outputs an impulse response vector.  As auto-correlation of an m-sequence
is near zero for any relative phase shift, the actual measured energy at a 
particular phase shift is related to the impulse response for that time delay.
This is the same principle used in spread-spectrum RAKE receivers such as are
used with GPS and CDMA.

The transmitter is designed to work only with the board in side A.  The 
receiver may be in side A or side B.  The boards may be standalone LFTX/LFRXs
or RFX daughterboards.

To use, the following script is installed into $prefix/bin:

Usage: usrp_sounder.py [options]

Options:
  -h, --help            show this help message and exit
  -R RX_SUBDEV_SPEC, --rx-subdev-spec=RX_SUBDEV_SPEC
                        select USRP Rx side A or B
  -f FREQ, --frequency=FREQ
                        set frequency to FREQ in Hz, default is 0.0
  -d DEGREE, --degree=DEGREE
                        set sounding sequence degree (2-12), default is 12,
  -t, --transmit        enable sounding transmitter
  -r, --receive         enable sounding receiver
  -l, --loopback        enable digital loopback, default is disabled
  -v, --verbose         enable verbose output, default is disabled
  -D, --debug           enable debugging output, default is disabled
  -F FILENAME, --filename=FILENAME
                        log received impulse responses to file

To use with an LFTX board, set the center frequency to 16M:

$ usrp_sounder.py -f 16M -t

The sounder receiver command line is:

$ usrp_sounder.py -f 16M -r -F output.dat

You can vary the m-sequence degree between 2 and 12, which will create
sequence lengths between 3 and 4095 (128 us).  This will affect
how frequently the receiver can calculate impulse response vectors.

The correlator uses an O(N^2) algorithm, by using an entire PN period
of the received signal to correlate at each lag value.  Thus, using a
degree 12 PN code of length 4095, it takes 4095*4095/32e6 seconds to
calculate a single impulse response vector, about a half a second.  One
can reduce this time by a factor of 4 for each decrement in PN code
degree, but this also reduces the inherent processing gain by 6 dB as
well.

The impulse response vectors are written to a file in complex float
format, and consist of the actual impulse response with a noise floor
dependent on the PN code degree in use.

There is a loopback test mode that causes the sounding waveform to be
routed back to the receiver inside the USRP:

$ usrp_sounder.py -r -t -l -F output.dat

The resulting impulse response will be a spike followed by a near zero
value for the rest of the period.

Synchronization at the receiver is not yet implemented, so the actual
impulse response may be time shifted an arbitrary value within the the
impulse response vector.  If one assumes the first to arrive signal is
the strongest, then one can circularly rotate the vector until the peak
is at time zero.

Johnathan Corgan
Corgan Enterprises LLC
jcorgan@corganenterprises.com
5/28/07