R_n in rake_mvdr_filters with real data

[valentin7121]

In example R_n in rake_mvdr_filters is set to R_n=sigma2_n * np.eye(mics.Lg * mics.M).

I guess that R_n is the noise covariance matrix and when processing a signal from real microphones it should be calculated based on the recorded signal. But I can't figure out how to correctly calculate R_n so that its dimensions match [mics.Lg * mics.M, mics.Lg * mics.M].

Maybe an explanation of the meaning of filter size Lg and beamformer delay would also help to better understand how to tune the beamformer.