A comparison of subspace adaptive filtering techniques for DS-CDMA interference suppression

We compare the performance of two subspace adaptive filtering techniques for suppressing multiple-access interference in a direct-sequence (DS) code-division multiple-access (CDMA) system. A military scenario is considered in which the processing gain is very large compared with the number of users. Also, communications is peer-to-peer so that power control cannot be used to solve the near-far problem. Two subspace techniques are considered. The first projects the received vectors on to an estimated signal subspace obtained by an appropriate eigen-decomposition. The second partially despreads the received signal. The dimension of the lower-dimensional subspace affects both the minimum mean squared error, and the response time to interference transients. There is a tradeoff between these two quantities which determines the optimum choice of dimension. This is illustrated numerically for a particular model in which the interferers turn on and off according to a two-state Markov chain. Simulation results are also presented which show that recursively updated algorithms are much more sensitive to large interference transients than block-oriented adaptive filtering algorithms