My PhD Research Activities

Some things that I have looked into in this area include

• Termination of Turbo Codes

  Termination of turbo codes pose a problem to designers as turbo codes are typically a block orientated system. Additional bits therefore are required to be sent accross the channel to terminate the two codes properly. Research I completed looked into using alternative methods of termination. The most successful method is not to terminate at all but instead to set the probability of all final states to be equal

• Variance Estimation for Turbo Code decoding

  Another problem associated with turbo codes is noise variance estimation. This is required as turbo code decoders utilise the MAP algorithm instead of the simplier Viterbi algorithm, which doesn't require this channel information. This work reviewed conventional variance estimation techniques. These techniques fail at low SNR (where turbo code decoders operate). We developed a data directed technique which gained the 0.1dB lost by using the conventional approach, numerical analysis of the variance estimator also confirms that we should achieve an improvement.

• Applications of turbo codes to multiuser receivers

  Turbo codes have been found to give large improvements in coding gain when applied to decoders. In this work we utilise this technique for coded DS-CDMA systems. We model the DS-CDMA channel as a convolutional (real) code and therefore have a serial concatenated convolutional code structure. We apply the turbo coding techniques to a number of different implementations and reach single user performance (effectively the best you can achieve). This result improves over previous solutions by at least 4dB but up to 8dB for large highly loaded systems.

  The unique design used random codes with systems up to K=27 users, and N=31 spreading lengths (spectrally efficient systems). This can be achieved for linear complexity.

• Analysis of the MAP algorithm and Turbo Code Decoding

  Turbo Code Decoding is yet to be fully explained by the coding community. The main block in a turbo code decoder is the MAP decoder. This work looks at numerically analysing the performance or noise variance reduction ability of the MAP decoder. Results from this work will be used to analyse the performance of Turbo Code Decoders and iterative multiuser receivers.

• Iterative Receiver for the Partial Response Channel

  Partial Response Signalling (PRS) is discussed in the literature as a means of transmitting at the Nyquist rate, with the detriment of introducing a large amount of intersymbol interference (ISI). In this work we apply our iterative multiuser receiver to mitigate the ISI generated by the PRS channel. Results, as expected, are close to the no ISI performance case, even for the difficult memory 5, E^2 PR4 channel.

• Iterative Detection of DS-CDMA using Antenna Arrays and FEC on Multipath Channels
  
  This work investigates a multi-point to point communication receiver system that accepts coded data modulated using either DS-CDMA or narrow-band BPSK, onto an Antenna Array.  We model the system with multiple signals arriving at the antenna array with direction of arrival (DOA) uniformly distributed. We describe an iterative receiver that improves the initial estimates from the antenna array, and therefore reducing the multiple access interference(MAI).
  
  Simulation and analytical results show that the performance approaches that when only one signal is incident on the array, even with a large number of users in comparison to the number of antenna elements and processing gain.