Decorrelating multiuser code-timing estimation for long-code CDMA with bandlimited chip waveforms

This paper addresses the problem of multiuser code-timing estimation for asynchronous uplink code-division multiple-access (CDMA) systems with aperiodic spreading codes and bandlimited chip waveforms. Two decorrelating code-timing estimation schemes, namely the frequency-domain least-squares (FLS) and frequency-domain weighted least-squares (FWLS) estimators, are developed. The two proposed estimators offer different tradeoffs between complexity and estimation accuracy. A critical step for decorrelating-based estimation is to decompose the received signal into subsignals of shorter duration. We discuss how to perform the decomposition to ensure improved identifiability and statistical stability of the proposed schemes. Due to a unique signal structure in the frequency domain, both the FLS and FWLS estimators admit efficient implementations that result in significant complexity reductions. The Crame´r-Rao bound for the estimation problem under study is derived and used as an assessment tool for the proposed estimators. Numerical results show that both of the proposed estimators can support overloaded systems (with more users than the processing gain) in multipath fading environments and significantly outperform a conventional technique based on matched-filter processing.