Ray-Based Versus Statistical Channel Models

Ray-based channel (RBC) models are physically linked to both array topology and the propagation environment. As a result, they form the basis of many standardized channel models and are commonly used in research. Nevertheless, their complexity means that many initial or analytical studies are still based on simpler statistical channel models. Since RBCs commonly assume a reasonable number of rays, the central limit theorem suggests that these channels may be approximately Gaussian. Hence, Rayleigh and Ricean approximations are obvious possibilities. Here, we explore these possibilities and present novel results demonstrating that simple statistical models can be used to closely approximate the signal-to-interference-and-noise ratio (SINR) performance of linear receivers with RBCs and highlighting the importance of user-specific correlation models.

I. Introduction

Ray-based channel (RBC) models offer several advantages over classical statistical models including a more direct physical motivation, a built-in dependence on array topology and validity over a wide range of frequency bands. As a result, RBCs form the basis of many standards models (e.g., [1]) and are widely used in simulations [2] and, more recently, in analysis [3]. Despite this adoption, RBCs are more complex and require more parameterization. For example, three dimensional RBCs require the generation of powers, elevation angles and azimuth angles for each of possibly hundreds of individual rays. Hence, they are less amenable to analysis. As a result, simpler channel models such as correlated Rayleigh and Ricean models [4], which only require spatial correlation matrices and line-of-sight vectors, are widely used.