An adaptive delay estimator for DS-CDMA

The Multiple Signal Classification (MUSIC) delay estimation algorithm is well known to be insensitive to the near-far effect in an asynchronous DS-CDMA system. Most MUSIC based estimators assume the channel is time invariant. In this paper we consider the performance of signal subspace tracking algorithms applicable to a time varying channel, and the effect of a time variant subspace dimension. In particular, we consider the Singular Value Decomposition (SVD) function, the Power Method algorithm incorporating Hotelling deflation (PMHd), and the Projection Approximation Subspace Tracking (PASTd) RLS-type algorithm. Simulations show that these algorithms can offer comparable performance, where the PASTd outperforms the latter functions in terms of computational load. This trait allows the use of the MUSIC algorithm for real-time delay acquisition schemes.