| This thesis investigates the iterative decoding of Turbo Code which consist in a parallel concatenation of two Recursive and Systematic convolutional Coders (RSC) separated by an interleaver.; The decoding of Turbo Codes suffers from great complexity. The maximum a posteriori (MAP) algorithm is among the two algorithms which allow the decoding of Turbo Codes, a simplified variant of the MAP algorithm, called the Log-MAP, is used in our research.; The first aspect of this work is to evaluate and analyze, throughout computations, the errors performance of the Turbo decoder using the Log-MAP algorithm for the constraint lengths K = 3 and 5, the total coding rates Rt = 1/2 and 1/3, and for a medium bloc size of lengths N = 196 to 3600 bits. Two types of channel are taken into consideration: an additive white gaussien noise channel and a fading channel based on Rayleigh model. Only the coherent modulation BPSK is considered in our simulations.; The drawback of using the MAP is the necessity to estimate the channel noise. Therefore, different methods for estimating the noise variance were examined and analyzed. The applied analysis is based on a probabilistic approach.; Finally, this work highlights the problem of trellises termination in Turbo Codes which is absolutely different from the one used for classical convolutional codes. Several alternatives for trellis termination are presented and compared. An alternative named MNoTail suppresses the necessity of sending tails. MNoTail is among the best techniques of termination regarding the error performance while its advantage is that no tail is added to the information sequence. |
