Differential spatial modulation : low complexity detection and improved error performance.

Abstract

Multiple-input multiple-output (MIMO) systems utilize multiple transmit and receive antennas in order to achieve a high spectral efficiency and improved reliability for wireless links. However, MIMO systems suffer from high system complexity and costs, due to inter-channel interference (ICI) at the receiver, the requirement of transmit-antenna synchronization (TAS), and the need for multiple radio frequency (RF) chains. Spatial modulation (SM) is a MIMO system which maintains a high spectral efficiency without suffering from ICI or TAS, while utilizing a single RF chain. However, the SM receiver requires full knowledge of the channel state information (CSI) to achieve optimal error performance, thereby increasing the receiver detection complexity. To overcome this, differential SM (DSM) has been developed which does not require CSI to perform detection. However, the maximum-likelihood (ML) detection for DSM results in excessive computational complexity when the number of transmit antennas is large, and suffers from a 3 dB signal-to-noise ratio (SNR) penalty compared to coherent SM. This dissertation aims to reduce the computational complexity of DSM, and mitigate the 3 dB SNR penalty. A generalized differential scheme based on SM (GD-SM) is proposed, which employs optimal power allocation to reduce the 3 dB SNR penalty. GD-SM divides a frame into a reference part and normal part. The reference part is transmitted at a higher power than the normal part, and is used to encode and decode the information in the normal part. Optimal power allocation is applied to the system, and the results demonstrate that at a bit error rate (BER) of 10−5 and for a frame length of 400, GD-SM is only 0.5 dB behind coherent SM. The frame structure of GD-SM and optimal power allocation is extended to conventional DSM (C-DSM). At a BER of 10−5, a 2.5 dB gain is achieved over C-DSM for a frame length of 400. Furthermore, the frame structure allows for easy implementation of quadrature amplitude modulation (QAM), which yields an additional gain in error performance. The use of QAM constellations is not possible in C-DSM. A simple, near-ML, low-complexity detector (L-CD) is proposed for DSM. The L-CD exploits the features of the phase shift keying, and amplitude phase shift keying constellations to achieve near-ML error performance, and at least a 98% reduction in computational complexity. The proposed detector is independent of the constellation size, and demonstrates a significantly lower complexity than that of current L-CDs.