Impact of Spatial Correlation on the Finite-SNR Diversity-Multiplexing Tradeoff

The impact of spatial correlation on the performance limits of multielement antenna (MEA) channels is analyzed in terms of the diversity-multiplexing tradeoff (DMT) at finite signal-to-noise ratio (SNR) values. A lower bound on the outage probability is first derived. Using this bound accurate finite-SNR estimate of the DMT is then derived. This estimate allows to gain insight on the impact of spatial correlation on the DMT at finite SNR. As expected, the DMT is severely degraded as the spatial correlation increases. Moreover, using asymptotic analysis, we show that our framework encompasses well-known results concerning the asymptotic behavior of the DMT.

💡 Research Summary

The paper investigates how spatial correlation among antenna elements influences the fundamental performance limits of multiple‑element antenna (MEA) or MIMO systems when the signal‑to‑noise ratio (SNR) is finite. The authors adopt the widely used channel model

 H = R_r^{1/2} H_w R_t^{1/2},

where H_w contains i.i.d. zero‑mean complex Gaussian entries, while R_t and R_r are positive‑definite Hermitian matrices that capture transmit‑ and receive‑side spatial correlation, respectively. The channel is assumed quasi‑static, unknown at the transmitter, and perfectly known at the receiver.

The mutual information for equal power allocation across N_t transmit antennas is

 I = log₂ det(I_{N_r}+η N_t H H^H)

with η denoting the average per‑receive‑antenna SNR. Outage probability, the key performance metric, is defined as

 P_out = Pr{I < R},

where R is the target spectral efficiency. Because an exact closed‑form expression for P_out is intractable, the authors derive a tight lower bound that can be evaluated efficiently.

The derivation starts by QR‑decomposing the uncorrelated matrix H_w = Q R. The diagonal entries |R_{l,l}|² follow chi‑square distributions with 2(N_r−l+1) degrees of freedom, while the off‑diagonal entries are i.i.d. standard complex Gaussian. By exploiting the identity det(I+XY)=det(I+YX) and the unitary invariance of the distribution, the mutual information can be upper‑bounded as

 I ≤ Σ_{l=1}^{t} log₂(1 + η N_t Δ_l),

where t = min(N_t,N_r) and

 Δ_l = Σ_{k=l}^{N_t} D_k² |R_{l,k}|².

The D_k are the singular values of R_t^{1/2}. The random variable Δ_l is a generalized quadratic form of Gaussian variables. Lemma 1 provides its probability density function as a finite mixture of Gamma distributions, with mixture weights a(l)_k and a(l+k)_1 that depend on the distinct singular values D_k.

Using the distribution of Δ_l, Theorem 1 gives two lower‑bound expressions for the outage probability. For the i.i.d. (uncorrelated) case (all D_k = 1) the bound reduces to a simple product of incomplete Gamma functions:

 P_out ≥ ∏_{l=1}^{t} Γ_inc(ξ_l, N_r+N_t−2l+1),

where ξ_l = N_t η