Performance Limits of MIMO Systems with Nonlinear Power Amplifiers

Abstract

Abstract-The development of 5G enabling technologies brings new challenges to the design of power amplifiers (PAs). In particular, there is a strong demand for low-cost, nonlinear PAs which, however, introduce nonlinear distortions. On the other hand, contemporary expensive PAs show great power efficiency in their nonlinear region. Inspired by this trade-off between nonlinearity distortions and efficiency, finding an optimal operating point is highly desirable. Hence, it is first necessary to fully understand how and how much the performance of multiple-input multipleoutput (MIMO) systems deteriorates with PA nonlinearities. In this paper, we first reduce the ergodic achievable rate (EAR) optimization from a power allocation to a power control problem with only one optimization variable, i.e. total input power. Then, we develop a closed-form expression for the EAR, where this variable is fixed. Since this expression is intractable for further analysis, two simple lower bounds and one upper bound are proposed. These bounds enable us to find the best input power and approach the channel capacity. Finally, our simulation results evaluate the EAR of MIMO channels in the presence of nonlinearities. An important observation is that the MIMO performance can be significantly degraded if we utilize the whole power budget.

Energy-Efficiency Maximization of Hybrid Massive MIMO Precoding With Random-Resolution DACs via RF Selection

“…Let us consider the well-known Bussgang's model [35] for the linear approximation of the introduced distortion of the quantization noise [36]. Given that Q(•) denotes a uniform scalar quantizer then for the scalar input s we have that…”

“…Remark: In this work we focus on modeling the nonlinearity effect of DACs. However, the approximation (7) could also include the nonlinearity effect of power amplifiers (PAs) [36]. This extension is left for future work.…”

“…The azimuth angles of arrival and departure are generated from the Laplace distribution with a standard deviation of 50 • . The number of Compute the EE for S i ∈ C via (36) and (42) 3: end for 4: Find S opt = arg max Si EE clusters set to N cl = 4 and the number of rays per cluster to N rays = 5. We assume digital combining is performed at the user equipment with N R = N s and the combiner matrix W is given by (22).…”

Robust Energy-Efficient Design for MISO Non-Orthogonal Multiple Access Systems

“…, K} and l = 1, 2. Let h l,k ∈ C N ×1 represent the channel vector from the BS to U l,k , which can be modeled as χ [49], where χ denotes the Rayleigh fading channel gain, d l,k is the distance between the BS and U l,k , and α represents the path loss exponent. For user pairing, we apply the clustering algorithm in [47] which is based on the channel correlation,…”

Optimising the Massive MIMO Downlink in the Presence of Power Amplifier Nonlinearities

“…Some papers model the effects of non-ideal PAs as either a fixed power additive complex Gaussian signal [6,34] or proportional additive complex Gaussian signal [1,4], added to the wanted transmit signal at the output of the PAs. Another approach used was to model the input to the PAs as a complex Gaussian signal input to an amplifier described by a nonlinear function [3,10,25]. In these cases, by using the Bussgang Theorem, the effect of nonlinearity can be modelled as the combination of a real multiplicative scalar and an additive noise term.…”