1. Page 73369, right column, in Algorithm 3, the definition of the beamforming vector: both terms inside the brackets are $e^{j\pi m\sin(\theta_2)}$; the first one should be $e^{j\pi m\sin(\theta_1)}$, i.e., $[\tilde{\mathbf{b}}]_m = [e^{j\pi m\sin(\theta_1)} + e^{j\pi m\sin(\theta_2)}]/\sqrt{2M}$.

2. Page 73370, right column, in the paragraph discussing Fig. 6 ("From the results, we also observe that ..."): "$\kappa = M\sin(\Delta\theta)$ for behavior 1" should be "$\kappa = M\sin(\Delta\theta/2)$ for behavior 1" (in the narrowband approximation; strictly $M\cos\theta_0\sin(\Delta\theta/2)$). This is consistent with the maximum TTD delay $M|\sin(\Delta\theta/2)|/W$ stated in Remark 2, and gives $\kappa \approx 22$ for the system parameters used in the simulations, as can also be inferred from Fig. 6(a).