Abstract:
The demand for secure and reliable communication is increasing which give rises to the evolution of modern and sophisticated digital communication systems. Most of the errors in the communication systems are introduced by the communication channel, and error correcting codes can be used to overcome this issue, thus bringing reliability in the communication link. In this thesis, the author has proposed two algorithms for constructing a parity check matrix (H) of two classes of LDPC code and the performance of these LDPC codes perform better than the existing ones. While designing such codes, special focus has been given to maximize and can predetermine the girth of H matrix. The first proposed algorithm is about constructing H matrix for Quasi-Cyclic low-density parity-check (QC-LDPC). Such type of codes is used in many number of applications as it is suitable for hardware implementations. In the proposed algorithm, the H matrix of QC-LDPC is constructed by sequentially maximizing the local girth for each block column of the matrix. The performance of the proposed algorithm is compared with other existing algorithms, such as modified array code (MAC), Sidara-Fuja-Tanner (SFT) and Magic square, and the results show that the proposed algorithm performs better than the existing ones in terms of bit-error rate and provides a higher girth. Another algorithm is proposed for constructing the H matrix with column weight 2, and it has the capability to determine the length of the girth. Such type of codes is less complex than other weights and has good potential performance in partial response channel. Moreover, it can also be used to design non-binary LDPC codes with small or moderate code lengths. The proposed algorithm carefully adds the bit and the check nodes in the Tanner graph while keeping the length of the girth until it fully expands to the required Tanner graph. However, the H matrix of the proposed algorithm can provide a higher girth when compared with progressive edge growth (PEG) and Bit-filling algorithms at same code length. The simulation results show that the proposed algorithm can also yield a better BER performance. In addition, we also show the effective of the girth parameter in the H matrix, which implies that if the length of the girth is large, it can yield a better performance in terms of BER when compared with the H matrix with a lower girth length. Furthermore, this thesis also introduces a technique to improve complexity of PEG algorithm, which can be used to construct the H matrix faster than the original one and provides the same H matrix