Anyone have success with USRP2 @ 20MHz to the host cleanly?

GNU Radio
1

Hi all,

Has anyone here had any success with receiving 16-bit samples from the
USRP2
@ 20MHz to the host without an onslaught of overruns? I did some
searching
on the list and some people suggested a RAID configuration, but this is
my
poor little laptop. Just wanted to get a general consensus.

  • George
2

On Nov 22, 2009, at 10:11 AM, George N. wrote:

Has anyone here had any success with receiving 16-bit samples from the USRP2 @ 20MHz to the host without an onslaught of overruns? I did some searching on the list and some people suggested a RAID configuration, but this is my poor little laptop. Just wanted to get a general consensus.

20 MHz @16bps is 80 MB/s right? That’s just not going to work on a
laptop hard drive without doing some major compression. I wrote a
gzip_file_sink at some point in the past but the libz compression was
too slow to even keep up with 8 MHz @ 16bps. You could try looking for
an alternate compression algorithm. I hear LZO is faster for instance.

IMO the best option for a laptop is an SSD. I do get about ~80 MB/s
sequential write with my Intel SSD. But 160 GB isn’t a lot if you’re
going to be doing 20 MHz sampling!

Dan

3

George N. schrieb:

Hi all,

Has anyone here had any success with receiving 16-bit samples from the
USRP2 @ 20MHz to the host without an onslaught of overruns? I did some
searching on the list and some people suggested a RAID configuration,
but this is my poor little laptop. Just wanted to get a general consensus.

  • George

Use a RAM-Disk to store the data:
sudo mount -t tmpfs -o size=500M tmpfs /ramdisk

Konrad

4

If latency and RAM usage is not an issue, increase the ethernet
packet_ring buffer (I’m normally using from 0.5 to 1 GB). I wrote a
patch some while ago and I think it is now in gnuradio, but I’m not
sure if it is in the python interface. This will probably help you
closer to the long term average of your laptop hard drive bandwidth.

juha

5

Thanks for the suggestions! I am digging the ramdisk idea…
unfortunately
the laptop only has 1G of memory, but I can still get a hardy 10 second
trace out of it :slight_smile: I liked the compression idea, Dan. I wonder if we
can
get something that can keep up with the data rate. It seems like a
challenge in its own.

  • George

On Sun, Nov 22, 2009 at 4:53 PM, Konrad Meier <