It's probably unreasonable to expect the ARM processor to deliver 56 Msps to the 9361 without underflowing. It's just not that fast. It's marginal even on a high performance desktop CPU. You could try implementing a noise source in the FPGA using RFNOC. That would almost certainly eliminate your bottleneck.
On Mon, Apr 10, 2017 at 9:27 AM John B. Wood <john.wood@nrl.navy.mil> wrote:
Hello, all. Perhaps I was a little too optimistic in attempting to use
the Ettus E310 as a broadband noise source. I'm using the GRC models
"Fast Noise Source" with uniform noise type connected to a "UHD: USRP
Sink" with a CF of 1 GHZ and sample rate set to 4 Msps. Running the
GRC-generated .py code on the E310's embedded Arm processor results in a
reasonably flat noise spectrum over ~3.5 MHz about the CF as displayed
on a spectrum analyzer. Attempting to increase the sample rate above 4
Msps to obtain a wider noise BW results in a whole lot of buffer
underruns (all those "U"s in the terminal window and a ragged noise output.
The advertised instantaneous BW of the E310 is 56 MHz but clearly the
available flat noise spectrum is no where near that. As always your
comment is most appreciated. Sincerely,
John Wood
U.S. Naval Research Laboratory
_______________________________________________
Discuss-gnuradio mailing list
Discuss-gnuradio@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
--
Very Respectfully,
Dan CaJacob
No comments:
Post a Comment