The throughput of some wireless multiaccess systems

We compute the throughput of some multiaccess wireless systems for delay-tolerant data communications, characterized by an infinite population of uncoordinated users accessing a common channel. The channel is affected by block fading, and the channel state is perfectly known to the receiver but unknown to the transmitters. To cope with multiaccess interference (MAI) and fading, the users employ retransmission of erroneously received packets. We consider unspread and randomly spread (code-division multiple-access (CDMA)) systems with decentralized (single-user) decoding and a system where the receiver employs joint multiuser decoding. The following conclusions can be drawn from our analysis: (a) unspread systems with packet retransmission outperforms CDMA systems with conventional detection, but are outperformed by CDMA with linear minimum mean-square error (MMSE) detection. (b) For all systems based on single-user decoding (SUD), there exists a threshold value of (E_b/N_o) below which the throughput is maximized by an infinite number of users per dimension transmitting at vanishing rate, and above which the throughput is maximized by a finite average number of users per dimension transmitting at nonvanishing rate. Moreover, as (E_b/N_o) increases, the optimal average number of users per dimension tends to one. In this sense, we say that the optimized systems "self-orthogonalize." (c) For the system based on joint multiuser decoding, a simple slotted ALOHA strategy is able to recover the throughput penalty due to fading in the limit for high (E_b/N_o), while an incremental redundancy (INR) strategy recovers the fading penalty for any (E_b/N_o).