Hi Hamed,
so, estimating anything from noise will be noisy if the estimation has
something to do with the estimated entity. (Compare, especially, the
Cramér-Rao Lower Bound: you can't bias-free estimate with zero error in
a noisy reception.) So, is it right if I assume that you measured or
calculated the SNR at your receiver (don't know from the top of my head
what SNR a noise stddev of 0.1 translates to, and the estimator variance
is unreasonably high in that scenario?
Best regards,
Marcus
On 5/3/23 16:01, Hamed Al-Zubi wrote:
Dears,
I am doing some analysis on channel state information of OFDM signals.
I used GNURadion rx_ofdm.grc (gnuradio/gr-digital/examples/ofdm at
main · gnuradio/gnuradio
<https://github.com/gnuradio/gnuradio/tree/main/gr-digital/examples/ofdm>)
for signal transmission and reception through a channel model block.
Then, the CSI has been extracted using Tag debug block.
After that, I used matlab to plot the phase of pilot signals from csi.
When I set the noise voltage on the channel block to zero, the phase
of each pilot signal is constant. However, when I set the noise
voltage to 100m, the phases of the pilot signals vary over time as
shown in Fig. (phase_pilot). I used a moving average filter to
mitigate the noise impact on the pilot phase. The results were better,
but still there is a phase drop as shown in Fig. (phase_pilot_filter).
As I read, the phase should be constant as long as the environment is
static. So, why there is a phase drop? how that can be fixed?
Thanks,
HZ
gnuradio/gr-digital/examples/ofdm at main · gnuradio/gnuradio
GNU Radio – the Free and Open Software Radio Ecosystem -
gnuradio/gr-digital/examples/ofdm at main · gnuradio/gn...
<https://github.com/gnuradio/gnuradio/tree/main/gr-digital/examples/ofdm>
