Search

Charge-trapping phenomena in radio-frequency (RF) power amplifiers (PAs) based on GaN high-electron-mobility transistor (HEMT) technology are understood to be responsible for the dynamic self-biasing behavior that leads to a seemingly intractable slow dynamic residual nonlinearity in communications applications. For this reason, and based on recent developments in the characterization and modeling of charge-trapping phenomena, in this article we demonstrate how the dynamic self-biasing behavior of GaN HEMT-based PAs can be characterized, modeled, and compensated. First, we describe a method for the accurate characterization of the capture and emission dynamics of charge-trapping phenomena using transient two-tone large-signal RF measurements. Then, we demonstrate that the accurate modeling of these phenomena is contingent on the capture process being described by a state-variable time constant, rather than a fixed near-instantaneous time constant as is typically assumed. Finally, we propose a fully analog electronic circuit that implements an approximation of the Shockley–Read–Hall (SRH) statistics-based physical model of charge trapping to compensate the dynamic self-biasing behavior of a 15 W GaN HEMT-based PA.

This paper presents a study of the linearizability of AlGaN/GaN HEMT based RF power amplifiers, RFPAs, and its relation with the active device trap activation energy. Based on the theory of thermally activated traps and on the experimental determination of the trap activation energy, we could show that despite different devices may exhibit traps with the same emission timeconstant at room temperature, their characteristic frequency may change significantly under nominal operation because of their temperature rise. And this was found to be key to explain the distinct linearizability performance of the tested devices because different stimulus dynamics excite the long-term memory effects imposed by traps with sensible different levels.

This paper provides theoretical and experimental evidence that, contrary to what is a widely reported belief, the capture time constant of GaN high-electron-mobility transistor (HEMTs) deep-level traps is not infinitesimally shorter than the modulation envelope time features of usual excitation signals. Instead, it can have a nonnegligible impact on their power amplification. A specifically conceived test bench, capable of measuring capture time constants at guaranteed invariant thermal dissipation conditions, revealed that the capture process can range from less than a microsecond up to a few tens of milliseconds. Furthermore, a theoretical justification based on the Shockley–Read–Hall statistics is provided to explain this widespread time constants’ behavior of deep-level traps. As a practical application example of these findings, the detailed characterization of the capture time constants performed in this paper proved to constitute a valuable tool in understanding the behavior of GaN power amplifiers (PAs) designed for pulsed radar signals.

This work is an overview of the current understanding of memory effects arising from AlGaN/GaN HEMT based RF power amplifiers. It reviews the most widely accepted physical sources of those memory effects and their impact on the amplifier's linearizability, to then discuss corresponding circuit design approaches to minimize the phenomena. This is illustrated by results of a power amplifier whose short-term and long-term nonlinear memory is compensated by an analog assisted digital pre-distortion scheme.

This work discusses the necessary modelling efforts needed to build nonlinear equivalent-circuit models that can produce consistent directcurrent (dc) and alternate-current (ac) characteristics of non-quasi-static devices such as GaN/AlGaN HEMTs, showing electro-thermal and/or trapping behavior. With this goal in mind, a measurement methodology addressing both pulsed dc I/V characteristics and pulsed S-parameters is presented and exemplifying results are shown.

This paper presents an explanation for the observed differences on the trapping time constants extracted from low frequency $Y_{22}$ and $I_{DS}$ -Deep-Level Transient Spectroscopy (DLTS) on GaN high electron mobility transistors (HEMTs). It is shown that the trapping time constant, not being a physical identity but a parameter of a model, can vary according to the conditions of extraction. This dependence is thus clearly seen in $Y_{22}$ and $I_{DS}$ -DL TS measurements, as they are made respectively under small-signal steady-state and large-signal transient conditions. A theoretical explanation supported by the Shockley-Read-Hall (SRH) model is proposed, showing how time-constants extracted from $I_{DS}$ -DLTS, i.e. under charge carrier emission, must be longer than the ones obtained from $Y_{22}$ , i.e. under equilibrium and thus where both emission and much faster capture processes play a role.

In this paper we propose an experimental method for the characterization and extraction of the time constants associated with the charge capture and emission processes of electron-trapping phenomena observed in radio-frequency (RF) power amplifiers (PAs) based on high-electron-mobility transistors (HEMTs). The method consists in the measurement of the transient response of a GaN HEMT-based PA to a series of large-signal two-tone RF excitations with increasing frequency separation, and the successive tracking of the gradual self-biasing experienced by the PA. With little requirements in terms of instrumentation and with applicability to fully assembled PAs, this simple method provides meaningful information on the dynamics of electron trapping close to the actual operating conditions of a GaN HEMT-based PA.

This work presents a novel approach to measure dynamic scattering (S-) parameters under transient pulsed bias conditions, aimed at the trapping characterization of microwave transistors. With the developed measurement technique, the trapping dynamics can be extracted from both the gate-to-source capacitance, C gs and the drain-to-source conductance, G ds , transients. This new technique provides a more comprehensive overview of the trapping mechanisms and can thus improve present models of microwave transistors. Experimental measurements of “transient pulsed” S-parameters on a commercial GaN HEMT are reported.

To address the issue of potential cholinergic neurotoxic effects in exposed biota, including sentinel or indicator species, a biomarker approach using cholinesterase (ChE) inhibition as effect criterion has been proposed. Waterborne metals and complex metallic effluents (e.g. from mining) also seem to affect ChE activity of numerous aquatic species, although there is debate on the likelihood of this effect. Considering that environmental exposure to mine effluents is a complex phenomenon, complicated by other factors (such as pH), we intended to simulate a uranium rich effluent, to which test organisms were acutely exposed. The objective of the present study was to quantify the effects of waterborne uranium on ChE activity of the freshwater organisms Daphnia magna (crustacean), Corbicula fluminea (mollusc), and Carassius auratus (fish), in order to evaluate the feasibility of using this biomarker to diagnose chemical stress induced by waterborne uranium, resulting from mining activities. The obtained data showed that, despite the large body of evidence pointing to inhibitory effects on cholinesterases caused by metals, uranium was not causative of significant ChE impairment in vivo for any tested organism, at concentrations that represent a worst-case contamination scenario. Considering the obtained results, cholinesterase inhibition is not suited to be integrated into a battery of biomarkers to diagnose uranium exposure. Although our study is strictly laboratorial, it has implications in the use and interpretation of this environmental biomarker in biomonitoring studies, especially of uranium-impacted areas.

Deep-level traps’ capture and emission time constants are known to impact the GaN HEMT models' ability to reproduce the device behavior when subject to RF modulated signals. Since deep level traps' capture time constants are usually much shorter than current pulse duration, they are usually neglected as assumed infinitesimally short. This work demonstrates, through basic semiconductor physics and experimental results, that such an assumption may not be verified, therefore justifying the effort of modeling such time constants as they may fall within the low-frequency components of modern communication signals.

This paper presents a simple method to extract the time constants associated with trapping effects observed in GaN HEMT devices. Instead of using specific and, sometimes, very expensive measurement systems, we propose a method that relies on common equipment present in all RF laboratories. The method uses measured information of a GaN power amplifier (PA) gathered from a VNA (small-signal gain collected for different gate-to source voltages) and a VSA (dynamic gain profiles obtained from a set of two-tone input signals with different separation frequencies). The paper starts by presenting the theoretical background of the proposed methodology followed by an experimental validation with a GaN HEMT PA.