Channel Matrices and Path Loss

I am trying to apply pathloss to channel matrices to use in a simulation. The paper I am referencing says:

"the channels are formulated as the product of large scale fading and small scale fading. The entries in the small scale fading matrix are randomly generated with complex zero-mean Gaussian random variables with unit covariance."

They are using PL = PL_0 + 10*C*log10(d/d_0) dB. The channel matrices are various sizes but are all complex normal, mean zero and unit covariance.

Can someone help me figure out how to interperet this? I am obviously doing something wrong.