The performance of turbo codes can often be improved by improving the weight spectra of such codes. Methods of producing the weight spectra of turbo codes have been investigated and many improvements were made to refine the techniques. A much faster method of weight spectrum evaluation has been developed that allows calculation of weight spectra within a few minutes on a typical desktop PC. Simulation results show that new high performance turbo codes are produced by the optimisation methods presented. The two further important areas of concern are the code itself and the decoding. Improvements of the code are accomplished through optimisation of the interleaver and choice of constituent coders. Optimisation of interleaves can also be accomplished automatically using the algorithms described in this work. The addition of a CRC as an outer code proved to offer a vast improvement on the overall code performance. This was achieved without any code rate loss as the turbo code is punctured to make way for the CRC remainder. The results show a gain of 0.4dB compared to the non-CRC (1014,676) turbo code. Another improvement to the decoding performance was achieved through a combination of MAP decoding and Ordered Reliability decoding. The simulations show a performance of just 0.2dB from the Shannon limit. The same code without ordered reliability decoding has a performance curve which is 0.6dB from the Shannon limit. In situations where the MAP decoder fails to converge ordered reliability decoding succeeds in producing a codeword much closer to the received vector, often the correct codeword. The ordered reliability decoding adds to the computational complexity but lends itself to FPGA implementation.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:577070 |
| Date | January 2013 |
| Creators | Rogers, Andrew John |
| Contributors | Ambroze, Marcel Adrian |
| Publisher | University of Plymouth |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/10026.1/1334 |
Page generated in 0.0023 seconds