Iteration Overlap and Dual-Sided LSOVA for Efficient Parallel Turbo Decoding Architectures

An efficient turbo decoder architecture is proposed by jointly exploring advancements in code design, decoding algorithms, and hardware implementation to achieve high-throughput and low-latency performance. At the algorithmic level, we introduce Dual-Sided Local-SOVA (DS-LSOVA), an enhanced variant of Local-SOVA that significantly lowers the complexity of extrinsic information computation while preserving decoding accuracy. To further improve implementation efficiency, iteration overlap interleavers are developed specifically for fully pipelined, iteration-unrolled UXMAP decoders, enabling full iteration overlap. This design reduces decoding latency and boosts hardware area efficiency by up to 22%. Importantly, these gains are achieved without sacrificing error-correcting performance: DS-LSOVA matches the Frame Error Rate (FER) of conventional max-Log-MAP decoding, while the proposed interleavers deliver FER results comparable to the Quadratic Permutation Polynomial (QPP) interleavers used in LTE.