The performance of cython_polar

[An00001]

Hello,
I have encountered an issue while running "my_project_with_aff3ct/examples/cython_polar/codec_polar.py." I noticed that when R=0.5 and R=0.59, the BER becomes 0 for SNR<0. However, all the literature I have read so far indicates that the BER should only drop to 0 when SNR>0. I am confused about this discrepancy and would appreciate your assistance.

The power in the “codec_polar.py” is obtained by calculating the variance of the codewords, and this is used to compute the variance (noise power) of the noise. However, based on my past observations, noise variance is typically calculated by setting Es=1 in other cases. I would like to know the differences between these two methods. Does the implementation of noise variance in these two ways affect the performance of BER?

[kouchy]

This is a contribution from @SubtleMuffin. Maybe he can answer you?