An Early Termination Strategy for Irregular LDPC Codes with Layered Decoding - Performance Evaluation and Implementation

Much attention has been paid to layered structures in the design of LDPC decoders. They promise fast convergence speed as well as reduced memory requirements. However, a correct stopping criterion (Hc checking) is difficult to implement in layered decoders, which results in unnecessary iterations. In this paper, we evaluate the error-rate performance of an hardware efficient early termination (ET) strategy under the normalized min-sum (NMS) algorithm and its improved versions. The irregular LDPC codes from IEEE 802.11n (WiFi) and IEEE 802.16e (WiMax) standards have been used for our investigation. From simulation results, it is shown that the conventional NMS algorithm with one optimal correction factor causes relatively high error floor. Therefore, we propose an improvement of the NMS algorithm by utilizing properties of degree-2 variable nodes. That improvement significantly lowers the error floor of the irregular LDPC codes with the ET strategy at high channel SNR values, which is confirmed by simulations. Finally, we employ the ET strategy in a WiMax decoder to reduce power consumption (around 50 % at a FER = 10(exp -4)).