Study On Performance And Decoding Implementation Of Turbo Code

Turbo codes is a new class of error-correcting codes that can approach the Shannon bound, it is considered as one of the most exciting and potentially important developments in code theory in recent years, its invention has changed the conventional design principles of the coding scheme from the attempt to increase the minimum Hamming distance of the code to the goal of reducing the number of code words with low Hamming weights. Turbo codes tap the potentialities of concatenated codes. For its excellent decoding performance, it has become the error-correcting coding scheme in 3rd mobile communication channel. So it is necessary to research Turbo codes' coding & decoding methods and its performance.This thesis investigates some problems of Turbo codes with emphasis on performance; its components code structures, modulation characteristics, decoding algorithm and the VLSI decoding implementation of serial and parallel. The main results and contents of this dissertation covers the follows:1. The development on the Turbo codes is outlined; the features and existing problems are pointed out. The coding theory of Turbo codes, decoding iteration, main decoding algorithm and improved decoding algorithm are expressed. Performances of various decoding algorithm are compared.2. The exciting performance of Turbo codes is explained in view of Hamming distance of Turbo codes. For component codes adopt recursion systematic convolution code with tailing Hamming distances, the Turbo code with interleaver size N can lower the code-error rate to 1/N multiple of convolution code. From the sequence correlation requiring of Turbo codes coding and decoding, correlation of sequence is proposed as a criterion to design interleaver. a good interleave scheme called clinodiagonal interleaver is obtained. Two methods to get high rate Turbo codes are investigated; the interleaver design of punctured Turbo codes is given. Decoding performances and complexities of these two methods are compared; the conclusion of punctured Turbo codes that can compromise between performances and complexities is got.3. The MAP decoding algorithm of Turbo codes requires a lot of middle variables recursion computations. This makes it be hard by VISL , so matrix implement of MAP algorithm is proposed, which simplifies the computation, speeds up the computation with the help of parallel computing and reduces the storage of middle variables. Because the matrix is a sparse matrix, the decoding can be simplified and its implementation by VISL can become easy.4. The MAP algorithm begins to calculate after receiving an entire frame data, so the delay is unbearable. The fixed delay MAP algorithm based on path conception in Veterbi decoding is put forwards, which only has forward recursion calculate and needn't an entire frame data, the delay of decoding is 5 to 10 times of the coding registers. It can implement real time decoding with little loss of decoding gain. It is an optimized algorithm.5. Based on improved MAX-LOG-MAP algorithm, serial circuit structure of Turbo codes decoding are given, methods of input data quantized and states & path metric are studied.