Implementation Trade-Offs of Soft-Input Soft-Output MAP Decoders for Convolutional Codes

Soft-input soft-output (SISO) maximum a-posteriori (MAP) decoders for convolutional codes (CCs) are an integral part of many modern wireless communication systems. Specifically, SISO-MAP decoding forms the basis for turbo decoders, as, e.g., specified for HSDPA or 3GPP-LTE, or for iterative detection and decoding in multiple-input multiple-output wireless systems, such as IEEE 802.11n. In this paper, we investigate the silicon-area, throughput, and energy-efficiency trade-offs associated with SISO-MAP decoders based on the algorithm developed by Bahl, Cocke, Jelinek, and Raviv (BCJR). To this end, we develop radix-2 and radix-4 architectures for high-throughput SISO-MAP decoding of CCs having 4, 8, 16, 32, and 64 states and present corresponding implementation results in 180 nm, 130 nm, and 90 nm CMOS technology. We validate technology-scaling rules and finally demonstrate the use of the presented trade-off analysis by identifying the key design parameters for parallel turbo-decoder implementations.