Abstract
When using full-rank MMSE combiner to detect signals in an asynchronous code-division multiple access(CDMA) system, the inverse of channel crosscorrelation matrix must be calculated. Because the matrix is related to spreading sequences, channel fading coefficients and user delay, it must be re-calculated each time a new symbol is transmitted. Due to this reason, the calculation complexity is not satisfactory. Later on, someone proposed reduced-rank MMSE combiner which can reduce the dimension used when calculating the inverse matrix by using MMSE criterion and the idea about projecting the signals onto the Krylov subspace. However, the weights of the proposed combiner still relate tightly with spreading sequence and channel realizations. Thus, the calculation complexity can not be greatly decreased because the weights still need to update with time. For further complexity reduction, asymptotic reduced-rank MMSE combiner is proposed.It has been shown that it can be used to replace real reduced-rank MMSE combiner when user number K and spreading gain N both tend to infinity. The performance measurements of the former, such as BER and Output SINR is close to the latter, and one important advantage of the former is that it can reduce the calculation complexity greatly. Because the above statements has been shown in direct-sequence CDMA (DS-CDMA) and multi-carrier CDMA (MC-CDMA), we will calculate the weights of the asymptotic reduced-rank MMSE in asynchronous MC-DS-CDMA, and different K/N and different channel fading distributions will be used in simulation. Finally, we will compare the performance of asymptotic reduced-rank MMSE combiner, real reduced-rank MMSE combiner and full-rank MMSE combiner, and then put forward the conclusion.