Multiuser diversity in delay-limited cellular wideband systems

We consider the uplink and the downlink of a multiuser wireless system with one base station and K user terminals. We model wideband transmission by considering M parallel subchannels, each of which is affected by fading. The fading processes in each subchannel are slowly time-varying with respect to the coding block length. Hence, in order to maintain given rate requirements for each user and each channel state, power control is used. We study the delay-limited achievable sum rate (throughput) versus the system E_b/N_O, under orthogonal and optimal signaling. We show that for both the orthogonal and the optimal schemes, in the limit of large K and finite M, the optimal allocation strategy consists of allocating each user to its best subchannel only. Hence, we are able to quantify the multiuser diversity gain by comparing the case K → ∞ with the single-user delay-limited case. Finally, we show that the limits of optimal signaling can be approached by relatively simple convolutional codes and iterative joint multiuser decoding with appropriate power control. The proposed scheme can be regarded as a practical version of the optimal successive decoding approach, that mitigates the error propagation due to the suboptimality of the user channel codes.