Reduced complexity decoding and relative performance of turbo codes

Abstract

The thesis has two main topics, both related to turbo codes. The first one deals with the design of reduced complexity decoding of turbo codes. In particular, two new decoders are designed using ideas from the M-algorithm: one to decode the constituent code and the other the entire turbo code. Both decoders are found to be suitable only for short turbo codes with information block size less than 300 bits. For these turbo codes the two decoders have a lower complexity with a small degradation in error performance when compared to the original decoder used to decode the turbo codes. Therefore they are recommended in applications where short turbo codes are used and a small processing delay is required. The second topic of the thesis is whether the turbo code is superior to any other code. A partial answer is found by comparing short BCH codes to turbo codes with similar parameters. It is found that the BCH codes are more attractive than the turbo codes in both decoding complexity and error performance. This implies that traditional block codes are better than turbo codes with similar parameters.