Efficient belief propagation decoding of polar codes: algorithms and architectures

Nowadays considering the transmission infrastructures of large data that travel at very high speeds, they must be transferred with considerable efficiency. Important results have been achieved almost at the limit of the maximum capacity of the transmission channel with different types of Error Correction Code (ECC), although they have never fully saturated it. The new generation polar codes, as demonstrated, are able, due to their recursion, to reach the maximum capacity of the channel with low complexity coding structures. Instead, in the information decoding process, the Belief Propagation, an iterative parallel decoding algorithm, has been studied, which allows to improve low latency performance. To eliminate the same latency times due to the iterative convergence procedure, the early stopping criteria have been studied and introduced. The objective of this thesis is the study of their performance, considering the complexity of the methods themselves; the introduction of a new early stopping method based on the Cyclic Redundancy Check (CRC) and finally a hardware implementation of the Frozen Bit Error Rate (FBER) criterion. We started with the analysis of the state of the art of the stopping criteria trying to compare the performances weighing them to the computational cost. It was decided to continue with the software implementation of the very low FBER method only. Furthermore, the criterion based on the CRC has been devised and implemented, which makes it possible to enter into synergy with the other methods given its construction. Finally, the FBER method was implemented in hardware considering an architecture optimized to minimize the cost of complexity by trying to maintain and improve its performance.