IEEE Transactions on Circuits & Systems II Express Briefs
An Efficient LDPC Decoder Based on Multi-Dimension Nonlinear Stochastic Computing
Kaining Han, Ziming Liu, Jianhao Hu
Abstract
Low-Density Parity-Check (LDPC) decoders face significant challenges in hardware complexity and routing congestion when targeting high throughput for next-generation communication systems. While stochastic computing offers a low-complexity alternative for fully parallel decoder implementation, it suffers from the hold-state problem in variable node updating, degrading performance and convergence. To address this issue, this paper presents a novel stochastic LDPC decoder based on a multi-dimensional nonlinear (MDN) stochastic computing unit. The proposed method formulates the update process as a multivariable optimization problem, achieving high accuracy. Implementation results show that the proposed decoder delivers comparable bit-error-rate (BER) performance while achieving 26% and 62% higher hardware efficiency for IEEE 802.3an and 5G NR LDPC codes, respectively.