Impact of Biasing Effects on Turn-On Switching Transients of GaN Power eHEMTs: Repeatability and Consistency

Abstract

Power electronic designers use I-V device characterization, with source meters and double-pulse testing, to choose devices and to validate the circuit layout of converters. Double pulse testing helps designers predict the performance and efficiency of the final converter, without the cost of high-power supplies and loads. However, unlike silicon devices, gallium nitride (GaN) enhancement-mode high-electron-mobility transistor (eHEMTs) are known to suffer from bias effects that impact the device characterization tests as they undergo temporary subtle changes in gate threshold voltage and on-state resistance as a function of their terminal voltages. This raises questions around how closely GaN double-pulse testing resembles continuous switching in a converter and how well source meter results relate to the I-V trajectory of a device that is continuously switching. This article investigates the impact of the bias effects on repeatability and measurement consistency of GaN eHEMT I-V characterizations. The devices evaluated are 650 V Schottky-gate and 600 V ohmic-gate GaN eHEMTs, while 900 V rated SiC mosfets are used as a control, due to their relative insusceptibility to historic terminal bias. The gate threshold voltage derived from source meter measurements is shown to depend on the meter's timings and pulse durations. The switching waveforms from double pulse testing are shown to be different to those of continuous switching, and this difference is a function of how gate and dc-link voltages are applied prior to switching. Continuous switching waveforms are shown to settle within a million cycles. Finally, a preconditioning approach for noncontinuous testing is demonstrated that employs a defined number of pulses to prebias the GaN eHEMT device, thereby causing measurement edges that resemble those of continuous-mode switching.