An investigation of Turbo Codes over Mobile Wireless Channels

Abstract

Turbo codes have been the subject of much research in recent years, producing results very close to the theoretical limit set by Shannon. The codes have been successfully implemented in satellite and video conferencing systems and provision has been made in 3rd generation mobile systems. These codes have not been used for short frame systems due to the delay at the decoder. In this thesis, comprehensive comparisons of the two common decoding algorithms are made, with reference to short frames. The effects of increasing memory size of component codes, frame sizes, utilising puncturing and errors in channel estimation are investigated over AWGN and Rayleigh fading channels. The decoder systems are compared for complexity as well as for equal numbers of iterations. Results show that less complex decoder strategies produce good results for voice quality bit error rates. Investigations are also made into the effects of errors in signal-to-noise ratio estimation at the SOVA turbo decoder, showing this decoding algorithm to be more resilient than Log-MAP decoders in published literature. The decoders are also tested over channels displaying inter-symbol interference. Channels include a time-invariant channel and three ETSI standard time-varying channels simulating indoor, pedestrian and vehicular situations, upgraded for more realistic Doppler effect. To combat these types of channels, a derivative of turbo codes, turbo equalisation is often used. To keep receiver delay to a minimum, decision feedback equalisation is used here. Results show that the combination can produce improvements in decoded results with increasing turbo iterations where ISI is low, but that iterative improvements do not occur under harsh circumstances. The combination produces much superior results compared with codes on their own under even the most extreme circumstances.