Multiple-rate codes based on serial concatenation and block Markov superposition transmission

This thesis proposed two improved channel coding schemes for digital communication systems, featuring in multiple code rates and low error floors. The first scheme is based on the well-known serially concatenated codes, while the other is based on the recently proposed block Markov superposition transmission (BMST) scheme. Serial concatenation of Hamming codes and an accumulator has been shown to achieve near capacity performance at high code rates. However, these codes usually exhibit poor error floor performance due to their small minimum distances. To overcome this weakness, we propose to replace the outer Hamming codes by product codes constructed from Hamming codes and singleparity-check (SPC) codes. In this way, the minimum distance of the outer code can be doubled, which is expected to increase the minimum distance of the serially concatenated code, thus improving the error floor performance. Moreover, the code rate can be adjusted by using different SPC codes. Three-dimensional EXIT chart is used for their convergence analysis and the found thresholds are shown to approach Shannon limit closely. Averaged ensemble distance spectra of the proposed codes is also calculated and compared with the original code. Simulation results show that the proposed codes can lower the error floor by two orders of magnitudes without waterfall performance degradation at short block length. Block Markov superposition transmission (BMST) is a recently proposed channel coding scheme for the construction of big convolutional codes from some short codes. This thesis investigates a new class of low complexity multiple-rate codes based on the recently proposed BMST scheme using the first order Reed-Muller (RM) and extended Hamming (EH) codes. Compared with the multiple-rate codes based on BMST of repetition and single-paritycheck (RSPC) codes, the new codes require much smaller encoding memories (and the number of interleavers used in the encoder) to achieve the same coding gain. Moreover, the decoding of BMST-RMEH codes has lower computational complexities (approximately half that of BMST-RSPC codes) and faster convergence speed