Chapter 13: DMT-Optimal Code Constructions

Learning Objectives

• State the Golden-code construction of Belfiore-Rekaya-Viterbo (2005) as a $2 \times 2$ full-rate space-time code over $\mathbb{Q}(j, \sqrt{5})/\mathbb{Q}(j)$, and prove that its minimum codeword-pair determinant satisfies $\delta_{\min} = 1/5$ independently of the input QAM size
• Define a cyclic division algebra $\mathcal{A}(F, K, \sigma, \gamma)$ of degree $n_t$ and explain, via the non-norm condition on $\gamma$, why every non-zero element of the algebra is invertible — hence every non-zero codeword matrix is full-rank
• Reproduce the Elia-Kumar-Pawar-Kumar-Lu-Caire 2006 construction of explicit DMT-optimal codes for any $(n_t, n_r)$, and state the non-vanishing-determinant (NVD) property as the algebraic fingerprint of DMT optimality at $r \to r_{\max}$
• Use the Tavildar-Viswanath definition of approximate universality: a code achieves the DMT under every fading distribution with a density bounded away from zero near zero — not just i.i.d. Rayleigh. Prove that CDA-NVD codes are approximately universal
• Relate the Perfect-codes family (Oggier-Rekaya-Belfiore-Viterbo 2006) to the CDA framework, state why Perfect codes exist only in dimensions $n_t \in \{2, 3, 4, 6\}$, and recognise the uniform-average-energy constraint that distinguishes them from arbitrary CDA codes
• Evaluate the practical complexity of decoding a CDA code via sphere decoding ($\approx M^{n_t^2/2}$ average, $M^{n_t^2}$ worst case), recognise why 5G NR uses low-dimensional Alamouti-style precoding instead, and anticipate the lattice-coded alternative of Chapter 17