Block based low complexity iterative QR precoder structure for Massive MIMO

Increasing data traffic has exposed the need for multi-user multi-antenna wireless communication systems. Exploiting the excess degree of freedom owing to additional antennas at the receiver with spatial multiplexing has led to ensure higher per-user data rates along with higher system capacities. Massive Multiple-Input Multiple-Output (MIMO) have receiving a lot of attention and it is an illustration of this demand for high rates, especially in the downlink. In this thesis, we study the Massive MIMO communication setup in the downlink data transmission. In addition, we attempt to solve the problem of system sum rate maximization with a constraint on computational complexity at the base station (BS). In contrast to uplink, in the downlink system, interference cancellation is required to eliminate signals intended for other co-users. As the user terminals have strict restraints on physical dimensions, processing capabilities and power availability are extremely limited (relative to the base station). Therefore, we study the solutions from the literature in which most of the interference cancellation is performed by the base station (pre-coding). Meanwhile, we maximize the sum rate and also consider the restrictions on the computational complexity at the base station. In this thesis, we also study and evaluate different conventional linear pre-coding schemes as well as how they relate to optimal structure of the solution which maximize the system performances. We also study one of the suboptimal pre-coding solutions known as Block-diagonalization (BD) applicable in the case where a receiver has multiple antennas and compare their performance. Finally, we notice that regular BD schemes are not deployed in Massive MIMO despite the promising results in terms of the system sum rate performance. The reason for this is that they are computationally heavy. In this thesis, we attempt to reduce the complexity of the BD schemes by partitioning channel matrix into corresponding square matrix. We make use of QR-based BD method to precode in a lower dimension channel matrix rather than computing all of them. The simulation of quasi-static is used. We generate the model and evaluate the performance in MATLAB. The simulated environment consists of single cell Massive MIMO with up to 512 antennas at the BS and 2 antennas per user. We have investigated two performance characteristics, system sum rate and computational complexity using various linear pre-coding schemes, including the proposed one. For the complexity, the simulation results show that the proposed BD pre-coding solution reduced complexity up to 90% compared to the other schemes. The number of flops required in the proposed solution is much less than the regular BD algorithm. It is shown that the proposed solution modifies the multiplexing order which reduced the complexity for the use in Massive MIMO. For the system sum rate, the results of the increase in the number of users with a fixed number of 128 and 512 BS antennas show that the proposed method achieved up to 24% and 35% respectively compared to the regular BD algorithm. Besides, it has achieved up to 60% and 65% respectively for a fixed number of 128 and 512 BS antennas compared to ZF algorithm. The results of the increase in the number of BS antennas with a fixed number of users, the proposed method achieved up to 20% and 60% compared to regular BD and ZF algorithms respectively. Although there were parts of the analytical results have a slightly lower sum rate compared to regular BD pre-coding, the overall results of the proposed BD method outperformed regular linear precoding. This is because the proposed low complex BD pre-coding solution is not a reliable quantity to reflect the true capacity. However, we observe that the proposed solution is an optimal pre-coding schemes and can be considered as an alternative in a more general framework for Massive MIMO with multiple antennas at the receiver.

Text FK 2021 58 - IR.pdf Download (1MB)

Actions (login required)

View Item