Analytical methods for CDMA systems with parallel interference cancellation: The large deviation approach

For a simple CDMA system, we compute the bit-error probability (BEP) with soft-decision parallel-interference-cancellation (SD-PIC). Instead of approximating the signal-to-noise ratio, we use a different measure to calculate performance. This measure is the exponential rate of the BEP, i.e., the limit of n(-)1 log(BEP) = -I, for the processing gain n--> infinity, where I depends only on the number of users. We show, using the rate as a measure, that SD-PIC improves the performance. The values of I follow as the solution of an optimization problem which can be calculated numerically. We use these results to derive the asymptotic behaviour of the rate for large k. We also derive results for the second order asymptotics of the BEP. Inclusion of second order asymptotics leads to excellent approximations.