Joint demapping and source decoding for multilevel modulation

In this paper we show how joint demapping for multilevel modulation and source decoding using the BCJR algorithm on a Markov source, reduce the source symbol error rate (SSER), when compared to the subsystems working independently. Simulations demonstrate that at low E _b/N_o values, our proposal performs better than systems that consider source compression and channel coding with short codelength if there is redundancy in the source information to be used. Iterative and joint iterative systems using LDPC, turbo codes or the turbo principle with a convolutional code, show excellent bit error rate performance and approach the Shannon limit as the code length is increased. However, when the size of the code is relatively short, their performance is poor at low E_b/N_o values, becoming worse than the uncoded signal at some points