[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Discuss-gnuradio] Re: Discuss-gnuradio Digest, Vol 67, Issue 15

From: sri ram
Subject: [Discuss-gnuradio] Re: Discuss-gnuradio Digest, Vol 67, Issue 15
Date: Fri, 6 Jun 2008 18:07:25 -0400

Thanks for the quick reply.  Please see inline responses.

Date: Fri, 06 Jun 2008 12:19:56 -0400
From: Greg Troxel <address@hidden>
Subject: Re: [Discuss-gnuradio] 802.11 on USRP+GNURadio
To: "sri ram" <address@hidden>
Cc: address@hidden
Message-ID: <address@hidden>
Content-Type: text/plain; charset=us-ascii

 Looking at the archives, I understand that I can receive 1 Mbps
 probe/beacon packets with code developed by BBN. I use their code and
 see packets at 1 Mbps from different nodes.
 However, I don't know of a way to have the USRP as a destination for a
 flow using standard packet generation tools like Iperf. So I setup a

We wrote code to inject the packets (802.11 frames) into a modified
NetBSD tap device that was an 802.11 interface rather than ethernet, and
then were able to use tcpdump.  All of this code is on the
acert.ir.bbn.com server.  We didn't address all the interactions with
the 802.11 state machines in net80211 and the GNU Radio implementation.
So you should be able to port this to Linux, but it's non-trivial.

Yes. I shall look into that in some time.

 UDP flow between a conventional 802.11bg AP and a Laptop. I capture
 the packets on air with the USRP and determine how many of the packets
 of this flow I am able to receive. But here, out of 1000 packets (1500
 bytes each) sent at 1 Mbps, the laptop is able to receive around 900
 packets but the USRP captures somewhere between 100 to 550 packets. I

If the laptop is only gettinng 900, that indicates something is wrong.
Are you sending these back to back as fast as you can?  I'd back off to
200 pps or something like that, and see what happens, and then gradually
increase the rate, watching CPU load.

I was running the GNURadio code and the WLAN card on the same machine. I performed another experiment where the WLAn card and iperf run on one laptop and the GNURadio on another laptop. In this case, the laptop gets all the packets. But the laptop running the GNURadio, receives only a fraction of the packets. I reduced the Iperf sending rate to 200kbps. Still things don't change.
Then I found out that just running the bbn_80211b_rx.py code causes the CPU load to increase.
top tells me, CPU(s): 79.0% us with a load average increasing very fast greater than 1.
This doesn't seem alright to me. Is this normal? Might this have to do with the OS (Ubuntu) or the linux Kernel version (2.6.15)?

 am wondering whether this makes sense. I thought that the BBN code
 would capture most of the packets provided the rate is 1 Mbps
 (disregarding probe packets from other APs). But this does not seem to

We did get most packets given a reasonable load

 I use gnuradio-3.1.2 on Ubuntu Dapper with a 2 GHz Intel core duo
 processor and 2 GB RAM.

We had 1.6 GHz Pentium M with 2 GB RAM (Thinkpad T43), with NetBSD
current from summer 2006 (pre 4.0).

I'd appreciate any suggestions on why this happens and what might be an alternative to try.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]