Re: [Discuss-gnuradio] "Error rate" block with USRP

[Tom Rondeau]

On Wed, Apr 16, 2014 at 10:31 AM, Azza <address@hidden> wrote:

Hi All,

I am working with GNU Radio Companion 3.6 on CentOS. I want to calculate the
BER, so I have been using the "Error rate" block.
First, I send a random QPSK-modulated signal through a simulated noisy
channel (AWGN). I calculate the BER using the function error:
BER=0.5*erfc(sqrt(snr)) and compare it to the output of the "Error rate"
block. I find almost same results.
Second, I send a random QPSK-modulated signal through one USRP and receive
it with another USRP. Here is the problem: contrary to expections, the
output of "Error rate" block is always 0.5, and it does not vary with the
variation of SNR.
You find attached the flowgraph of this second scenario.
<http://gnuradio.4.n7.nabble.com/file/n47625/gnu-ber.png>
I am wondering if this block should not be used with USRPs to calculate the
BER. If not, any suggestions?
Any help is appreciated. Thank you.

Regards,
Azza

Azza,

First, you should take out the throttle. With the USRPs in the loop, you're going to have conflicts with the rates that will cause problems.

Second, at the very least, you need to synchronize everything through the hardware. The constellation decoder doesn't do any timing, phase, or frequency synchronization. You'll also need an AGC block to scale the received samples before anything else.

Tom