Adaptive feedforward/feedback detection architectures for wideband CDMA

Detectors for high data rate wideband code-division multiple access (CDMA) networks must contend with both multiple access interference (MAI) and intersymbol interference (ISI). Our reverse link detector suppresses the MAI and ISI with a feedforward filter (FFF) which linearly processes the samples at the output of parallel chip matched filters (CMFs), and a feedback filter (FBF) which linearly processes detected symbols. We propose a fully-connected (FC) architecture whose synergy arises from sharing FFF and FBF contents among users to enable efficient multiuser recursive least-squares (RLS) adaptation towards the minimum mean-square error (MMSE) coefficients. Simulations validate the theoretical BER analysis and demonstrate that the FC detector trains in several hundred symbols, avoids severe error propagation, and supports 8 simultaneous 2 Mb/s QPSK users in an asynchronous CDMA system with a processing gain of 8, a delay spread of 1.25 /spl mu/s, an average SNR of 15 dB, with uncoded BERs less than 10/sup -4/ and 10/sup -3/ with 97% and 99.3% probabilities, respectively.