Abstract
Coherent detection of asynchronous Code-Division MultipleAccess (CDMA) data transmissions over a Rician fading channel is considered in the context of a multipoint-to-point communication system, where a centralized receiver that is assumed to have knowledge of the signature signals of the system uaers, including the arrival times of the former, observes a superposition of the specular and faded signal components of each of the users in additive noise, The channel itself is assumed to be non-dispersive, and with no fading memory, i.e., the random attenuations and phase shifts experienced by the users’ transmissions over different bit intervals are assumed to be independent of each other, rendering them inestimable. It turns out that this channel is equivalent to a fictitious CDMA-AWGN (Additive White Gaussian Noise) channel from the point of view of optimal detection [I]; it may be shown that the optimum decision rules over the two channels as well as the statistical characterization of the sufficient statistics in each case parallel each other. Unlike the optimum AWGN multiuser detector however, the optimum faded detector is unimplementable using a Viterbi algorithm. This motivates the derivation of the polynomial- complexity faded decorrelator for this channel. Detector performance in a multiuser faded environment may be adversely affected by both the interferer specular (or known) and faded (or unestimatable) signal components. The performance limiting effect of the former on conventional detection as well as the ability of fading channel strategies to withstand such interference has been studied earlier [I]. The issue of the limitations on detector performance, if any, due to interfering fading is addressed here. To this end, we introduce the fading susceptance and fading resistance measures; the former as a measure of whether degradations in detector performance due to interfering fading are so great so as to prevent them from being compete tive with optimum detection strategies over single-user channels, and the latter as a measure that captures the ability of detectors to withstand such interference These asymptotic measures characterize detector performance in regions where the fading of the interfering users as opposed to their specular energies, is the dominant impediment to detection. An analysis of the conventional detector’s performance in the multiuser faded environment reveals that fading interference is capable of limiting detector performance in a manner similar to specular component interference bringing out the hitherto unrecognized performance limiting effect of fading interference in Rician fading CDMA channels. The AWGN niultiuser detectors (optimal and decorrelating), which are designed for the CDMAAWGN channel, while sub-optimally resistant to specular interference, pay a penalty for ignoring the presence of fading in that they are also found to be susceptible to fading interference. Thus we demonstrate that, even in environments where specular interference is marginal, both of the above detectors are incapable of competing with detectors of isolated transmissions, if the fading of an intefering user dominates the background noise. The faded detectors, both optimal and decorrelating, are however found to withstand such interference. In light of their previously demon46 strated resistance to specular interference, and the additional computational considerations, our results make the case for the use of the faded decorrelators for multiuser detection over asynchronous Rician faded CDMA channels. The plots of Figure 1 are an illustiation of the irnplications of our results for detector Bit-Error Rates (BERs) in realistic as opposed to asymptotic environments. We observe that, over a two-user channel, even with a weak interferer specular signal, the first user of the faded decorrelator alone exhibits an exponential decay in BER with increasing Signal-to-Noise Ratio (SNR) with both the conventional detector and the AWGN decorrelator forced by the non-zero, fixed interferer fading, to approach error floors
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have