Design of an Area-Efficient High-Throughput Shift-Based LDPC decoder

An area-e±cient high-throughput shift-based LDPC decoder architecture is proposed. The specially designed (512, 1,024) parity-check matrix is e®ective for partial parallel decoding by the min-sum algorithm (MSA). To increase throughput during decoding, two data frames are fed into the decoder to minimize idle time of the check node unit (CNU) and the variable node unit (VNU). Thus, the throughput is increased to almost two-fold. Unlike the conventional architecture, the message storage unit contains shift registers instead of de-multiplexers and registers. Therefore, hardware costs are reduced. Routing congestion and critical path delay are also reduced, which increases energy e±ciency. An implementation of the proposed decoder using TSMC 0.18 m CMOS process achieves a decoding throughput of 1.725Gbps, at a clock frequency of 56MHz, a supply voltage of 1.8V, and a core area of 5.18mm. The normalized area is smaller and the throughput per normalized power consumption is higher than those reported using the conventional architectures.