Broadband Applications with USRP

Hello Everyone,

I am interested in developing broadband applications (~30MHZ BW) using
the
USRP and I have some questions:

  1. What is the largest bandwidth I can process using a flow graph with
    relatively large complexity (e.g. 16QAM demodulator simple lowpass, 2 x
    mixers), any USRP and a computer with arbitrarily large processing
    power. I
    just want a rule of thumb approximation.

  2. I know it’s probably not the right place to ask, but do you think my
    application can be implemented in either Labview, or MATLAB USRP
    interfaces?

  3. If it is not possible at all, do you think Nutaq, PicoSDR is the
    right
    hardware, considering that it has modules implemented on the FPGA?

  4. Are there any papers/applications notes etc in the literature that
    you
    are aware of that implement broadband (30 MHZ BW) systems using the
    USRP?

Thank you

Eleftherios(Lefteris) Kampianakis
Electronics and Computer Engineer
PHD Candidate and Researcher at Sensing Computing Communications Group
(SGCC)
Department of Electrical Engineering
University of Washington
3927 Adams Lane, NE, Mercer Court D805B, 98105
website: http://staff.washington.edu/ekampian/
http://users.isc.tuc.gr/~ekabianakis/
mail: [email protected]

You can try my DVB-S2 modulator to do some transmit benchmarking.

With the test flow graph, to send a 30 MHz wide signal (which is
actually
typical for real DVB-S2 uplinks), you’ll need to set the symbol rate to
30 / 1.2 or 25 Msymbols/sec. Since the modulator uses 2X oversampling,
the sample rate will be 50 Msps.

You’ll need an SDR that can operate at that sample rate with an
interface that can keep up. A B2x0 could work if you have a good
USB3.0 controller.

The CPU bottleneck on the DVB-S2 flow is the FEC (both LDPC and BCH)
generation. The net bit-rate at 25 Msyms/s with the test flow graph
will be 87 Mbps.

In general, receive is usually much worse CPU wise. Again, the FEC can
take an enormous amount of CPU cycles.

best regards,
Ron

Hi Lefteris,
in addition to what Ron wrote:

On 12/04/2014 10:44 PM, Lefteris Kampianakis wrote:

Hello Everyone,

I am interested in developing broadband applications (~30MHZ BW) using the
USRP and I have some questions:

  1. What is the largest bandwidth I can process using a flow graph with
    relatively large complexity (e.g. 16QAM demodulator simple lowpass, 2 x
    mixers), any USRP and a computer with arbitrarily large processing power. I
    just want a rule of thumb approximation.
    If your computer is arbitrarily fast: the maximum sampling rate you can
    get out of the USRP.
    That would be something like 25MHz for N200/N210 @16bit, 50MHz @8bit,
    200MHz out of the X300/X310,
    and something mainly depending on your USB3 controller for the
    B200/B210.
  2. I know it’s probably not the right place to ask, but do you think my
    application can be implemented in either Labview, or MATLAB USRP interfaces?
    Yes, but in Matlab they will only be offline processing applications.
  3. Are there any papers/applications notes etc in the literature that you
    are aware of that implement broadband (30 MHZ BW) systems using the USRP?
    Yes, there are :slight_smile:

This one is not involving GNU Radio, but it’s impressive nevertheless,
since it’s 128 times a 30.72MHz sampling rate[1].

Then there’s a load of papers you’ll find when searching IEEE Xplore for
“USRP gnuradio”, and one of the nice things about GNU Radio (as much as
about the USRPs) is that sample rate is just a setting – if you can set
the desired sample rate, and your algorithms can deal with that rate,
you’re fine.

There’s inherently high demand for bandwidth in radar applications [2],
so choosing radar as a topic is a good approach for finding papers
describing USRP usage at high sample rates.

Greetings,
Marcus

[1]

[2] KIT - CEL: Communications Engineering Lab

Lefteris Kampianakis wrote in post #1164056:

Hello Everyone,

I am interested in developing broadband applications (~30MHZ BW) using
the
USRP and I have some questions:

  1. If it is not possible at all, do you think Nutaq, PicoSDR is the
    right
    hardware, considering that it has modules implemented on the FPGA?

The PicoSDR platform is well suited for real-time, high bandwidth
applications. The PicoSDR can be, at first used as a radio peripheral
but later on, you can migrate GNU Radio processing blocks into the FPGA,
to offload the GNU Radio flowgraph, thus deacreasing the load on the CPU
side. You also have a specific flavor of PicoSDR which is the PicoSDR
2x2-E which comes with an embedded i7 blade right next to the FPGA blade
making it a good development platform for hybrid processing
applications.

Regards,

JB