Abstract
Given an ordered J group partition of the K simultaneously transmitting users of a CDMA channel, a sequential group detector consists of J group detectors that are connected sequentially. The j/sup th/ group detector uses the decisions from the previous j-1 group detectors and cancels the inter-user interference from those users before it makes joint decisions for the j/sup th/ group. This successive interference cancellation scheme was introduced by M.K. Varasani (see IEEE Trans. Inf. Theory, vol.41, no.4, July, 1995) for the Gaussian CDMA channel. This paper consists of extending that idea to the frequency-selective Rayleigh fading (FSRF) CDMA channel. The two group detectors (I and II) for the FSRF-CDMA channel are considered as the basic building blocks. The resulting sequential group detectors can be regarded as members of two distinct classes (each class parametrized by the ordered partition) of multiuser detectors that satisfy a wide range of complexity constraints. A succinct indicator of the average BER over high SNR regions for the FSRF-CDMA channel is defined via the asymptotic efficiency. In this work, upper and lower bounds on the asymptotic efficiency for the two sequential group detectors are derived. Minimax criteria under which these detectors are optimal are specified. A numerical example illustrates the vast improvements achievable by the sequential group detector based on the group detector II.