Novel architectures for squares, and sums of squares, of cross-correlations of bipolar sequences with applications to CDMA

This paper describes novel architectures for computing squares, and sums of squares, of cross-correlations of bipolar sequences. Both metrics are used in improved channel decoder metrics for CDMA systems, but also appear in other signal processing applications. The key idea is to compute outer products of each code vector with itself, and formulate the metrics so as to avoid explicit computation of cross-correlations. The expensive squaring unit is also eliminated. A detailed complexity analysis is carried out in which we demonstrate the advantages of these architectures. They also offer easier updates when there are incremental changes in the system. We also describe a few variations and special cases including short codes and asynchronous users, and point out further enhancements.