Studies on Energy and Spectral Efficiency for Massive MIMO

Abstract

Energy consumption in cellular communication is increasing rapidly. There is a demand for energy efficient wireless communication systems due to depletion of conventional energy sources i.e. fossil fuels. At the same time, the total system capacity is expected to increase from current generation cellular communication system 4G to next generation system 5G. Massive MIMO is one of the possible disruptive ways for the 5G wireless communication system. In this technology base station (BS) equipped with a large number of antennas and it supports high antenna array gain as well as the uncorrelated channel. Due to uncorrelated channel vector, the expected cross-correlation between two channel vectors tends to zero. Thus, there is no effect of interference in uplink and downlink scenario which causes high SNR. In single-user (SU) massive MIMO system, the total L number of antennas across the BS communicate with a single user. The closed form expression for energy and spectral efficiency trade-off by considering the transceiver power consumption has been derived. For small spectral efficiency (SE), the energy efficiency (EE) linearly dependent on SE. In contrast for sufficiently large SE, the system requires more power to transmit which may not be the energy efficient. The effect of channel gain G on optimal EE has been analyzed in our present work. For the small cell size, the average channel gain is very large and vice versa. If the channel gain is very large, then the EE is insensitive to the power amplifier (PA) power efficiency. The multi-user (MU) massive MIMO system serves a multiple number of users U. In the case of perfect channel state information (CSI), the power transmitted by each user terminal (UT) can be suppressed proportional to the number of BS antennas L. Whereas in the case of imperfect CSI, the BS has to estimate the channel matrix using the pilot sequence transmitted along with the data. Here the transmit power can be scaled inversely proportional to square root of L. The present work suggests the lower bound on achievable data rate with the detection techniques zero forcing (ZF) and maximal ratio combining (MRC).