Composite CDMA - A statistical mechanics analysis

Code Division Multiple Access (CDMA) in which the spreading code assignment to users contains a random element has recently become a cornerstone of CDMA research. The random element in the construction is particular attractive as it provides robustness and flexibility in utilising multi-access channels, whilst not making significant sacrifices in terms of transmission power. Random codes are generated from some ensemble, here we consider the possibility of combining two standard paradigms, sparsely and densely spread codes, in a single composite code ensemble. The composite code analysis includes a replica symmetric calculation of performance in the large system limit, and investigation of finite systems through a composite belief propagation algorithm. A variety of codes are examined with a focus on the high multi-access interference regime. In both the large size limit and finite systems we demonstrate scenarios in which the composite code has typical performance exceeding sparse and dense codes at equivalent signal to noise ratio.

💡 Research Summary

The paper introduces a novel “composite” spreading‑code ensemble for Code Division Multiple Access (CDMA) systems that blends the two classical paradigms: sparse spreading and dense spreading. In the proposed construction each user’s spreading vector of length N is composed of a fraction p of dense components (drawn from a standard random Gaussian ensemble) and a fraction 1‑p of sparse components (a small number of non‑zero chips placed at random). By varying p the ensemble continuously interpolates between the pure dense case (p = 1) and the pure sparse case (p = 0). The authors investigate the performance of this ensemble both analytically in the thermodynamic limit (N, K → ∞ with load β = K/N fixed) and numerically for finite‑size systems.

The analytical treatment relies on the replica method from statistical mechanics. Assuming replica symmetry (RS), the authors derive closed‑form expressions for the free energy, the average bit‑error rate (BER), and the spectral efficiency (capacity) as functions of the system load β, the signal‑to‑noise ratio (SNR), and the mixing parameter p. The RS solution reveals that, especially in the high multi‑access interference (MAI) regime (β ≈ 1 or larger), there exists a range of intermediate p values for which the composite code outperforms both the pure sparse and pure dense codes. The improvement stems from a synergistic effect: the dense part supplies enough overall signal power to keep the detection problem well‑conditioned, while the sparse part reduces the effective interference by limiting the number of overlapping chips.

To complement the asymptotic analysis, the paper proposes a “composite belief propagation” (BP) algorithm tailored to the mixed structure. Standard BP works efficiently on sparse factor graphs but becomes intractable on fully dense graphs. The composite BP therefore treats the sparse sub‑graph with ordinary message passing, while the dense sub‑graph is handled by a mean‑field approximation that aggregates the contributions of many weakly coupled variables into a single effective field. This hybrid scheme retains the linear‑time scaling of sparse BP (O(N p) operations per iteration) while capturing the essential dense‑code statistics. The algorithm’s convergence properties are examined as a function of p, β, and SNR. In simulations the algorithm typically converges within 15–20 iterations, a modest increase over pure sparse BP but far fewer than would be required for a full dense BP implementation.

Extensive Monte‑Carlo experiments are reported for several representative scenarios. With load β = 1.0 and SNR = 10 dB, the BER reaches its minimum around p ≈ 0.4, achieving roughly a 30 % reduction compared with either extreme (p = 0 or p = 1). When the load is increased to β = 1.2, the optimal mixing shifts slightly toward a larger sparse fraction (p ≈ 0.3), confirming that the sparse component becomes more valuable as interference grows. Moreover, the composite code exhibits an “effective SNR gain” of about 1–2 dB: for a target BER the required SNR is lower than that of pure codes by this margin. The authors also compare computational costs, showing that composite BP’s runtime is comparable to that of a pure sparse implementation while delivering the performance benefits of the dense component.

The discussion acknowledges the limitations of the replica‑symmetric assumption. In extreme regimes (very high load or very low SNR) the RS solution may become unstable, hinting at the need for replica‑symmetry‑breaking (RSB) analysis. Likewise, the composite BP can experience convergence difficulties when the dense fraction is too large, suggesting that more sophisticated damping or adaptive temperature schemes could be beneficial. The paper outlines future work directions, including an RSB treatment, the development of variational BP variants, and the exploration of adaptive p‑selection based on real‑time channel conditions.

In summary, the study demonstrates that a carefully balanced mixture of sparse and dense spreading codes can deliver superior performance in CDMA systems, especially under heavy multi‑user interference. The combination of a rigorous statistical‑mechanics analysis and a practical composite belief‑propagation decoder provides both theoretical insight and a feasible algorithmic pathway for next‑generation wireless networks where massive connectivity and energy efficiency are paramount.