Noncoherent Interference Alignment: Trade Signal Power for Diversity Towards Multiplexing

Abstract

This paper proposes the first known universal interference alignment scheme for general $(1\times{}1)^K$ interference networks, either Gaussian or deterministic, with only 2 symbol extension. While interference alignment is theoretically powerful to increase the total network throughput tremendously, no existing scheme can achieve the degree of freedom upper bound exactly with finite complexity. This paper starts with detailed analysis of the diagonality problem of naive symbol extension in small $(1\times1)^3$ networks, a technique widely regarded as necessary to achieve interference alignment with insufficient diversity. Then, a joint bandpass noncoherent demodulation and interference alignment scheme is proposed to solve the diagonality problem by trading signal power for increased system diversity, which is further traded for multiplexing improvement. Finally, the proposed noncoherent interference alignment scheme is extended to general $(1\times{}1)^K$ cases and is proven to achieve the degree of freedom upper bound exactly. Simulation results verify the correctness and powerfulness of the proposed scheme and show significant degree of freedom improvement compared to the conventional orthogonal transmission scheme.