From: [email protected]main.invalid
[[email protected]main.invalid] on
behalf of Martin B. (CEL) [[email protected]]
Sent: Wednesday, November 14, 2012 12:29 PM
To: [email protected]
Subject: Re: [Discuss-gnuradio] Paper: DCSK chaotic modulations with GNU
On Tue, Nov 13, 2012 at 07:31:30PM +0000,
[email protected] wrote:
Feedback and comments are welcome. If somebody thinks this has a
place in the main repo, let me know what remain to be done.
Out of curiosity, how were you able to determine Eb/N0 in Fig. 3? Could
you use the SNR measurement blocks in GNU Radio?
Basically, we measured everything “by hand”, using the standard
block in GNU Radio 3.5 (plus some modifications in the “simple
framer”). So, no, we didn’t use the SNR estimation blocks in GNU
Radio (if I remember well, these blocks were not merged in GNU Radio
when we did these experiments, last year). We had the setup described
in section 5 of the paper (2 usrp, a noise generator with a mixer).
For each point of the graph, we measured the noise power alone, the
signal power alone, then the bit error rate when mixing the noise and
the signal. With some matlab code we converted the SNR to Eb/N0 to
generate the Fig 3 (instead of using a SNR estimation technique). We
did this because we wanted to actually measure the total performance
loss of the whole transmission system (modulation, implementation,
sync. recovery, usrp) and compare the results with the theoretical
limits. Moreover, I have no idea how far is a usrp transmit channel
from a pure theoretical AWGN one.
Other things that would be nice to test would be to compare this
“hand made” SNR measures with SNR estimation techniques, both in a lab
setup and in a real-world outdoor environment.
– Julien Olivain.