Measurement techniques for the characterization of radio frequency gallium nitride devices and power amplifiers

The rapid growth of mobile telecommunications has fueled the development of the fifth generation (5G) of standards, aiming to achieve high data rates and low latency. These capabilities make use of new regions of spectrum, wider bandwidths and spectrally efficient modulations. The deployment of 5G relies on the development of radio-frequency (RF) technology with increased performance. The broadband operation at high-power and high-frequency conditions is particularly challenging for power amplifiers (PA) in transmission stages, which seek to concurrently maximize linearity and energy efficiency. The properties of Gallium Nitride (GaN) allow the realization of active devices with favorable characteristics in these applications. However, GaN high-electron mobility transistors (HEMTs) suffer from spurious effects such as trapping due to physical defects introduced during the HEMT growth process. Traps dynamically capture and release mobile charges depending on the applied voltages and temperature, negatively affecting the RF PA performance. This work focuses on the development of novel measurement techniques and setups to investigate trapping behavior of GaN HEMTs and PAs. At low-frequency (LF), charge dynamics is analyzed using pulsed current transient characterizations, identifying relevant time constants in state-of-the-art GaN technologies for 5G. Instead, at high-frequency, tailored methods and setups are used in order to measure trapping effects during the operation of HEMTs and PAs in RF modulated conditions. These RF characterizations emulate application-like regimes, possibly involving the control of the device’s output load termination. Therefore, an innovative wideband active load pull (WALP) setup is developed, using the acquisition capabilities of standard vector-network-analyzers. Moreover, the implications of performing error-vector-magnitude characterizations under wideband load pull conditions are studied. Finally, an efficient implementation of a modified-Volterra model for RF PAs is presented, making use of a custom vector-fitting algorithm to simplify the nonlinear memory operators and enable their realization in simulation environments.