I need to generate a signal of dimension (K*1) where K is the number of users, for my system model [y = sqrt(Pu)HWS + N] with the following :

1. each user power is sqrt(Pu) and is equal to 36dbm.

2. the signal matrix is i.i.d random matrix with E{SS^H} = I

So, i have generated it in the following way:

Pu_dbm = 36; % input power of each user symbol in dbm

N_users = 10; % number of users

Pu_linear = 10.^((Pu_dbm/10)-3);

S = sqrt(Pu_linear)*complex(randn(N_users,1), randn(N_users,1));

Is there anything wrong in the implementation, because i see that i do not get the required statistics (SINR) or performance when i put this inside my model. So there has to be some problem in the above signal generation.