Finite-Length Scaling Of RA Codes On The BSC

As a class of capacity-achieving ensembles,Repeat-Accumulate codes(RA codes) are widely used in numerous practical applications due to their fast convergence and low complexity in encoding and decoding.They thus become the focal point of error-correcting code.Finite-length scaling law is a powerful tool to analyse the performance of moderate size capacity-achieving ensembles with iterative decoding.In this thesis,we use such scaling law to analyse the performance of finite-length RA codes,assuming that the transmission takes place over the BSC(binary symmetric channel) and Belief Propagation decoding is used with a finite number of iterations.The main contents are as follows.Firstly,we show how to re-derive the scaling law and scaling parameter for RA codes transmitted over the BSC.Secondly,we demonstrate how to use the scaling law as an estimate tool for(3,3) RA codes.Finally,we compare the predictive results on the frame error probability with simulation results to show that both of them match very well.