USRP hardware purchase

GNU Radio
1

Hi,

For those who have experience with the hardware, please see if my list
of components is adequate for a single host. The intent is to be able to
build a 802.11 (b/g) network.

  1. One USRP board package (including motherboard, enclosure, 2 RF
    cables, USB cable, power supply etc)
  2. Two RFX2400 daughter boards.
  3. Two “2400-2480 MHZ ISM band” antennas

Did I miss anything?

Thanks.

  •     David
2

I’ve got to go with Dan on this one, there is no chance of 802.11 with
the USRP and GNU Radio. However, we are working on something
802.11-like with the in-band signaling work and the MAC we are building.

I think a big first step for the platform is getting a solid contention
based MAC protocol, which the platform lacks as a whole. From this,
some 802.11-like MAC can be built, and other protocols.

To give you a heads up Dan, we are working on doing carrier sense in the
FPGA and supporting it with the in-band signaling work. It’s actually
what we’re working on at this second. Building it in the FPGA avoids
the FPGA->host->FPGA latency. The FPGA is going to read carrier sense
related parameters from registers, and the host will set a flag in the
packet if it wants the FPGA to carrier sense before transmitting it. I
think very time sensitive functionalities like this need to be pushed in
the FPGA… though we’re also considering building it in the host too
so we can evaluate the tradeoffs.

  • George
3

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The USRP has a maximum bandwidth of about 16 MHz (using 8-bit samples
instead of the usual 16), which is not even enough to cover one
802.11b/g channel (22 MHz Nyquist). Thus you’ll have a hard time
decoding b/g packets at faster than 1,2 Mbit rates (BBN has done some
work facilitating the recovery of these packets).

On top of that, an (extremely, I think) optimistic lower bound for
latency in a GNU radio system is over 1 millisecond from the RX antenna
through the FPGA over the USB to the computer to your software back over
the USB and out the TX antenna. This makes it reallllllly hard to get
the 10 microsecond turnaround required for an 802.11b ACK, and carrier
sense is meaningless with such a latency.

Also, see Tom R.'s email of 12/05/2006 for more comments, but the
short of it is that you’re unlikely to be able to get any sort of
reasonable interoperation between the USRP and commercial 802.11b/g
hardware.

  • -Dan
    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)
    Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGw67By9GYuuMoUJ4RAqhgAKDENetmsE0mIa9dw3JyE787xo8KMQCfX+Pf
Odcvnn/HpAQ8PDahyfKt0Fo=
=Hre7
-----END PGP SIGNATURE-----

4

On Wed, Aug 15, 2007 at 06:56:17PM -0700, Dan H. wrote:

On top of that, an (extremely, I think) optimistic lower bound for
latency in a GNU radio system is over 1 millisecond from the RX antenna
through the FPGA over the USB to the computer to your software back over
the USB and out the TX antenna. This makes it reallllllly hard to get
the 10 microsecond turnaround required for an 802.11b ACK, and carrier
sense is meaningless with such a latency.

I think 100us is possible if you’re mindful of the buffer sizing.

I concur that 10us is not going to happen on the host :wink:

Eric