Far-Field-Based Nonlinear Optimization of Millimeter-Wave Active Antenna for 5G Services

Abstract

This paper presents the design and characterization of millimeter-wave (28/38 GHz), circularly polarized (CP) active antennas, suitable for future 5G services. By augmenting the modeling capabilities of commercially available nonlinear CAD tools, the active antenna design can simultaneously optimize figures of merits for both radiation and nonlinear (NL) performance. The radiating part is computed and optimized layoutwise by means of an artificial neural network (ANN), suitably trained off-line. For the NL design purpose, the harmonic neural network (HNN) of the antenna is subsequently implemented as a standard circuit component, to include the antenna behavior at all the harmonics of its NL regime. This allows avoiding time-consuming electromagnetic (EM)-simulations in the harmonic balance optimization loop. In this way, a well-defined interface between the antenna and the amplifier can be avoided. To demonstrate the effectiveness of the design approach, one active integrated antenna (AIA), consisting of a class AB amplifier feeding a CP patch at 38 GHz, has been fabricated with the standard 0.1- $\mu \text{m}$ AlGaAs-InGaAs pHEMT technology and extensively measured with respect to both the electrical and radiation performances. To reduce fabrication costs, hybrid integration of the antenna and amplifier connection is used, and the antenna is incorporated into the main PCB.