Interference cancellation and turbo code are two major techniques to enable high data rate and to improve capacity in CDMA-based third generation mobile communication systems. The former significantly reduces most of the multiple access interference (MAI) while the latter provides very strong error correcting capability against residual interference and background noise....[ Read more ]

Interference cancellation and turbo code are two major techniques to enable high data rate and to improve capacity in CDMA-based third generation mobile communication systems. The former significantly reduces most of the multiple access interference (MAI) while the latter provides very strong error correcting capability against residual interference and background noise.

Investigations have been made on the possible configurations of combining the interference cancellation and turbo decoding, namely the concatenated and the integrated approaches. The concatenated approach is to perform all the multistage interference cancellation iterations first before the output is fed into the turbo decoder. As both interference cancellation and turbo decoding are iterative multi-stage algorithms, it is possible to integrate these two multi-stage techniques and an integrated interference cancellation and decoding algorithm is proposed.

In the proposed integrated multi-stage algorithm, each stage consists of an interference cancellation iteration and a turbo decoding iteration. We use the Soft-in/Soft-out Viterbi Algorithm (SOVA) in the turbo decoding iteration and apply the soft reliability value in the soft-decision interference cancellation iteration. Variations on the system were proposed and verified for further enhancement of the system performance.

We evaluated the performance of this integrated approach and showed that it has significant performance improvement over the usual concatenation approach. The performance gain of the proposed algorithm over the concatenation approach was found to be fairly robust under imperfect power control and are quite independent of the number of users and frame sizes.