I’ve been experimenting with a pair of USRP2s using gnuradio with a
custom communication system at 10MHz sampling rate at both the
transmitter and receiver). I’ve noticed that with minimal processing of
the received stream it seems to (occasionally) skip samples in bursts as
if an Ethernet packet was dropped. I have real time scheduling enabled
and the maximum usage on any given processor I’ve seen is about 40-50%
so it’s not a CPU limitation.
Is it normal to loose samples sent/received from the USRP and if so is
there any way around it?
Based on other forum posts (http://www.ruby-forum.com/topic/212514) it
seems like dropped packets can happen but I would think in many SDR
applications this would be a huge problem. I wouldn’t think that I would
be dropping any packets with each USRP2 directly connected via cat5e to
an Ethernet interface on my PC. I’m not really sure where to start
tracing down where the dropped samples are though so I’m looking for any
I transmitted a periodic stream at 10MHz (with interpolation of 10)
which was viewed on a software scope and clearly isn’t missing any
samples. The same stream at 10MHz (decimation of 10) on the receiver was
marked in the data with the same period as the transmitter.
Theoretically the marking position shouldn’t change relative to the
pattern over time. Triggering (with a software scope) on the mark shows
that its position within the period shifts leading to the conclusion
that some of the samples are missing. No “S”'s are printed to the
My hardware configuration is:
-2 USRP2s connected by a short sma cable
-Each USRP2 is direct connected with cat5E to a network interface on the
-10MHz shared reference clock provided to both USRPs
-6 core Xeon 3.3GHz processor
-30-40% average CPU usage and no large spikes in usage
-Ethernet buffer sizes/USRP2 are currently at defaults.
Let me know if you have any idea as to what might be causing the missing