Turbo coding and decoding for signal transmission and recording systems

In this dissertation, the application of concatenated codes with iterative decoding to signal transmission and recording systems is investigated. A general overview of parallel and serial concatenation of interleaved codes is given. The corresponding iterative (turbo) decoding algorithms are discussed along with the soft-input soft-output (SISO) component decoders. It is shown that a parallel concatenated convolutional code, or turbo code, is a special case of a serially concatenated convolutional code.; Different interleaver designs and some of their properties in a concatenated coding system are discussed. The distance spectrum of a turbo code is a function of the component encoders and the interleaver, and we show how particular interleaver mappings give low weight codewords. We introduce a new class of interleavers, called ST-random interleavers, and compare their performance, via simulation, to other interleavers.; A new method to improve the performance of a given turbo code based on the distance spectrum is introduced. Dummy bits is inserted into the most error prone bit positions in the encoder, and then removed in the decoder. Bit error rate performance improvement of one order of magnitude was observed.; Turbo decoding for interleaved codes on partial-response channels with additive white Gaussian noise (AWGN) is addressed. A system employing multiple single parity-check code as an outer code is introduced. This system offers similar performance to previously proposed systems, but significantly lower complexity allowing for faster decoding, which makes it an attractive scheme for magnetic recording systems.; A novel approach to estimate the performance of high rate interleaved codes on the partial response channel is introduced. We introduce a new method to incorporate a runlength constraint in the serial concatenation of an interleaved outer code and a precoded channel, while maintaining the benefits of turbo-equalization. The same method is used to improve the performance of error correcting codes over correlated fading channels with a time-delay constraint. We show analytically improvements in performance for block coding over the Gilbert-Elliott channel model matched to the land mobile channel. Then we show, via simulation, how the performance is improved for convolutional codes over a Rayleigh fading channel.