In the presence of channel fluctuation, rate adaptation is one way to maintain the quality of the link at a desired level. This is especially important in ad hoc wireless networks, where temporary channel fluctuations might create frequent needs for rerouting that would result in severe overhead and adversely affect the performance. We study an unusual method of passive rate adaptation in which some bits are dropped at the receiver end of a link. The symbol-error probability decreases as some bits are dropped. In terms of the distortion for a realtime analog signal, the tradeoff is between more reliable detection of fewer bits and less reliable detection of more bits. Our scheme achieves smaller distortion for a certain region of signal-to-noise ratio (SNR) values when compared with the original scheme without rate adaptation. Two examples, uniformly spaced, uncoded pulse amplitude modulation and quadrature amplitude modulation, are studied and compared for both a Gaussian channel and a Rayleigh fading channel. We conclude that our scheme is more suitable to use in a fading channel than in a Gaussian channel. We also verify that our scheme has a larger applicable region of SNR values when a nonuniform constellation is used, since the important bits are given additional protection.

In this paper, we consider a synchronous code-division multiple-access (CDMA) system with a multiuser receiver. All users are assumed to have symmetric signature sequences, but the presence of a subset of the users is unknown to the receiver. We first calculate the signal-to-interference ratio (SIR) in this environment for the matched-filter receiver, the decorrelating receiver, and the linear minimum mean-square error (MMSE) detector. We then identify the user capacity for a single-class system, and the effective bandwidth for a multiple-class system. The result is compared to the case of random sequences and of optimum sequences. For symmetric sequences, the effective bandwidth cannot be expressed by a scalar as in , because two constraints have to be satisfied simultaneously to satisfy the SIR requirement. We introduce a two-dimensional (2-D) vector notion of effective bandwidth with and without unknown users. For both the decorrelator and the MMSE detector, the user capacity is 1 when all users are known to the receivers and is reduced to (1-N/L) when N users are unknown (with L the processing gain). The performance of these three linear detectors, with and without unknown users, is compared.