Abstract-The open problem of calculating the limiting spectrum (or its Shannon transform) of increasingly large random Hermitian finite-band matrices is described. In general, these matrices include a finite number of non-zero diagonals around their main diagonal regardless of their size. Two different communication setups which may be modeled using such matrices are presented: a simple cellular uplink channel, and a time varying inter-symbol interference channel. Selected recent informationtheoretic works dealing directly with such channels are reviewed. Finally, several characteristics of the still unknown limiting spectrum of such matrices are listed, and some reflections are touched upon.
I. PROBLEM DESCRIPTIONConsider a linear channel of the formwhere x is the N K × 1 zero-mean complex Gaussian input vector x ∼ CN (0,y is the N × 1 output vector, and z denotes the N ×1 zero-mean complex Gaussian additive noise vector z ∼ CN (0, I N K ), which is independent of x and H N . Accordingly ρ = P K is the transmitted signal-tonoise ratio (SNR). In addition, the N × N K channels transfer matrix H N is defined bywhere {a i , b i , c i } are statistically independent 1×K random row vectors with independent identically distributed (i.i.d.) entries a i,j ∼ π a , b i,j ∼ π b , and c i,j ∼ π c . For simplicity, we assume that the power moments of the entries for any finite order are bounded. Finally, α, β ∈ [0, 1] are constants. The normalized input-output mutual information of (1) conditioned on H N (also known as the Shannon transform)where λ i (H N H † N ) denotes the ith eigenvalue of the Hermitian five-diagonal matrix H N H † N . Furthermore, denoting the indicator function by 1{·},is the empirical cumulative distribution function of the eigenvalues (also referred to as the spectrum or empirical distribution) of H N H † N . Fixing K and assuming that F HN H † N (x) converges almost surely (a.s.) to a unique limiting spectrum, it can be shown that the expectation of (3) with respect to (w.r.t.) the distribution of H N converges as well. This is since (3) is uniformly integrable due to the Hadamard inequality and the bounded power moment assumption, and hence the a.s. convergence implies convergence in expectation [1]. In Section II it will be realized that if the channel H N is known at the receiver and its variation over time is stationary and ergodic, then the expectation of (3) w.r.t. the distribution of H N is the per-cell sum-rate capacity of a certain cellular uplink channel model. In another setting (see Section II), the same expectation may be interpreted as the capacity of a certain time variant inter-symbol interference (ISI) channel, assuming again that the channel is known at the receiver.
A. Analytical DifficultyMany recent studies have analyzed the asymptotic rates of various vector channels using results from the theory of (large) random matrix (see [2] for a recent review). In those cases, the number of random variables involved is of the order of the number of elements in the matrix H N , and self-av...