A comparison of long versus short spreading sequences in coded asynchronous DS-CDMA systems

The performance of turbo-coded asynchronous direct sequence code division multiple access (DS-CDMA) using long and short spreading sequences is compared by both analysis and simulation. For coded systems with a conventional matched filter (MF) receiver, three analytical methods with different complexity are compared: the standard Gaussian approximation, the improved Gaussian approximation (IGA), and the density function approach. It is shown that while the standard Gaussian approximation is fairly accurate for the long sequences, it is too optimistic for the short sequences. For the short-sequence systems, the IGA gives an accurate estimate for the performance with much less complexity than the density function approach. The analysis shows that for either the additive white Gaussian noise (AWGN) channel or the flat Rayleigh fading channel and a MF receiver, there is a degradation in the average performance of the turbo-coded short-sequence systems compared to the long-sequence systems due to the fact that the cross-correlations are not time-varying. However, the short-sequence systems are amenable to the use of an interference suppression technique designed to minimize the mean square error. Such a minimum mean square error (MMSE) receiver in the turbo-coded system is shown to outperform the long-sequence system with the MF receiver, especially when there is a near-far problem, as previously observed in a convolutionally-coded system. Finally, similar results are obtained by computer simulations for the turbo-coded CDMA systems on a frequency-selective Rayleigh fading channel.