Title:High Throughput Polar Decoding Using Two-Staged Adaptive Successive Cancellation List Decoding

Abstract:Polar codes are the first class of capacity-achieving forward error correction (FEC) codes. They have been selected as one of the coding schemes for the 5G communication systems due to their excellent error correction performance when successive cancellation list (SCL) decoding with cyclic redundancy check (CRC) is used. A large list size is necessary for SCL decoding to achieve a low error rate. However, it impedes SCL decoding from achieving a high throughput as the computational complexity is very high when a large list size is used. In this paper, we propose a two-staged adaptive SCL (TA-SCL) decoding scheme and the corresponding hardware architecture to accelerate SCL decoding with a large list size. Constant system latency and data rate are supported by TA-SCL decoding. To analyse the decoding performance of TA-SCL, an accurate mathematical model based on Markov Chain is derived, which can be used to determine the parameters for practical designs. A VLSI architecture implementing TA-SCL decoding is then proposed. The proposed architecture is implemented using UMC 90nm technology. Experimental results show that TA-SCL can achieve throughputs of 3.00 and 2.35 Gbps when the list sizes are 8 and 32, respectively, which are nearly 3 times as that of the state-ofthe-art SCL decoding architectures, with negligible performance degradation on a wide signal-to-noise ratio (SNR) range and small hardware overhead.