Here are two indirect paths to model the antenna cross polarization
a) Using the "Combined" Spatial correlation matrix property option
Since it is the antenna cross-polarization effect that is the cause of the different path filters, if you can express the Spatial Correlation as a combined matrix that includes this effect, the actual path gains per link (4 links for a 2x2 system) would thus be different across the diagonal and cross-diagonal.
This includes a separate additional task in determining the correlation matrix, based on the user's parameters.
In general, the effect of the correlation matrix specification leads to different "average" path gains across the different links. Refer to the "Examine Spatial Correlation Characteristics of 2-by-2 Rayleigh Fading Channel" example on the comm.MIMOChannel doc reference page:
An example of a model that offers the cross-polarization effect directly as a property is the TDL channel model in 5G Toolbox (nrTDLChannel), which in essence computes this "combined" correlation matrix for the comm.MIMOChannel component. See here for more info on nrTDLChannel:
b) If the user has the actual path gains and delays per link, then they could create 4 SISO channels and process the signal through them and handle the receive end combining on their own.
Best Answer