Fabio Filicori

Envelope Tracking of an RF High Power Amplifier With an 8-Level Digitally Controlled GaN-on-Si Supply Modulator

This paper presents an envelope tracking (ET) transmitter architecture based on the combination of a novel 3-bit (N = 3) supply modulator and digital predistortion (DPD). The proposed power converter is based on a direct digital-to-analog conversion architecture that implements the binary-coded sum of N isolated dc voltages, allowing the synthesis of an output waveform with L = 2N voltage levels, with a binary distribution in the range ΔV = VM - VO (maximum voltage VM, offset voltage Vo). This solution provides a better voltage resolution VS = ΔV/(2N-1) with respect to typical multilevel switched-sources topologies (VS = ΔV/N). The improved voltage resolution enables the correction of the residual discretization error in the ET transmitter by means of DPD of the RF signal without the need of an auxiliary linear envelope amplifier. The proposed ET solution has been tested with an L-band 30-W lateral-diffused MOS RF high power amplifier (RF HPA) with 1.4- and 10-MHz long-term-evolution signals. In these conditions the converter demonstrated 92% and 83% efficiency, respectively, whereas the congregate efficiency of the transmitter are 38.3% and 23.9% at 5.5 and 1.9 W of average RF output power, respectively. These performances correspond to an improvement of 17.2 and 17.9 points for the power-added efficiency of the RF HPA and to 13.4 and 13 points of improvement for the efficiency of the entire transmitter with respect to fixed bias operation.

A Double-Pulse Technique for the Dynamic I/V Characterization of GaN FETs

Standard dynamic characterization methods based on periodic narrow-pulse low duty-cycle excitation waveforms provide suboptimal I/V curves when used along with GaN field effect transistors (FETs), due to complex nonlinear charge trapping effects. Thus, a double-pulse technique for the dynamic characterization of GaN FETs is here presented. The double-pulsed I/V characteristics are shown to be not only isothermal but also corresponding to a fixed charge trapping state.

A Three-Port Nonlinear Dynamic Behavioral Model for Supply-Modulated RF PAs

We propose a three-port nonlinear dynamic behavioral model for supply-modulated power amplifiers (PAs). The proposed model not only accounts for the radio frequency (RF) input-output relationship, but also for the interaction between a modulated voltage supply, the RF output power and the supply current. The model is based on a modified Volterra formulation which accounts for the dynamic deviations with respect to a quasi-static model. The frequency-domain kernels of the proposed model are directly extracted from measurements performed with a low-frequency-extended large-signal network analyzer on an RF hand-set PA. The model is validated under random multitone modulated RF input and supply. The presented technique allows for the independent control of the RF and the supply ports. As such, it allows a separate description of both the dynamic contribution of the RF modulated input and of the dynamic supply voltage. The proposed model shows an improvement with respect to a quasi-static approach in predicting the RF output, the supply current, as well as the power-added efficiency.

Pre-pulsing characterization of GaN PAs with dynamic supply

Nonlinear charge-trapping observed in the electrical characteristics of GaN FETs can introduce distortion in GaN-based power amplifiers (PA), especially in supply-modulated (envelope tracking) transmitters. A measurement approach is developed for large signal characterization of GaN-based PAs operated with dynamic bias supplies for efficiency enhancement. A new pre-pulsing technique is introduced which forces the active device to operate in trapping and thermal states close to those found in the actual application. The characteristics obtained with this technique are shown to give an accurate description of the PA performance. The measured data are used for the direct computation of pre-distortion functions for the linearization of a 10-GHz Envelope Tracking (ET) 12-W GaN MMIC PA for amplitude-modulated pulsed radar transmitters. The demonstrated measurement method can be also exploited for the identification of PA behavioral models, which take into account trapping effects.

Iso-thermal and iso-dynamic direct charge function characterization of GaN FET with single large-signal measurement

A fast and simple method for the direct characterization of nonlinear charge functions of electron devices is presented. The input and output transistor ports are simultaneously excited through single-tone sources at different frequencies and calibrated large signal waveforms are measured by means of an advanced NVNA-based setup. Proper choice of the two frequencies guarantees an almost complete coverage of the voltages domain in a single and very fast measurement set and allows the extraction of the charge functions by direct integration of currents in the frequency domain, since, contrary to other methods, the measured waveforms are both iso-thermal and iso-dynamic (i.e. at fixed charge trapping status). The method is validated by characterizing the gate charge function of a 5W 8×125μm GaN FET and implementing a simple table-based model of the transistor input port. Very good results are achieved by comparison with large-signal measurements under conditions different than the ones used for the characterization.

Charge-Controlled GaN FET Modeling by Displacement Current Integration From Frequency-Domain NVNA Measurements

We propose an efficient procedure for the extraction of a charge-controlled nonlinear model of a 1-mm gallium nitride on silicon carbide field-effect transistor (L = 0.25 μm) from nonlinear vector network analyzer acquisitions. A fast, single-shot measurement technique is described, in which the two device-under-test (DUT) ports are excited by single-tone sources at carefully selected tone frequencies, acquiring calibrated waveforms at the on-wafer DUT ports with an almost complete coverage of the voltages domain. The gate and drain charge functions identification is executed by the integration of the displacement currents in the frequency domain. A suitable approach for separating the conductive and displacement drain current components is provided. The presence of thermal self-heating and charge trapping phenomena is empirically evaluated, and accounted through an equivalent voltage approach. Experimental validation is provided at 2.5 and 5 GHz for a continuous-wave excitation, and at 2.5 GHz for a two-tone excitation.

Large-signal GaN transistor characterization and modeling including charge trapping nonlinear dynamics

A double-pulse technique for the I/V characterization of GaN-based transistors is adopted for the nonlinear modeling of a 0.25 μm AlGaN/GaN on SiC FET. Experimental validation is provided by means of large-signal PA performance prediction both at low-frequency, in order to outline the role played by the resistive drain current source, and at microwaves. Improved prediction accuracy is demonstrated with respect to the use of standard single-pulse I/V characteristics.

Two-Input Nonlinear Dynamic Model Inversion for the Linearization of Envelope-Tracking RF PAs

We present an algorithm for the real-time inversion of a two-input behavioral model applicable to supply-modulated radio-frequency (RF) power amplifiers (PAs). This approach includes the nonlinear dynamic effects on the RF output due to the mutual interactions between the modulated supply and the RF input. The model inversion consists of an iterative procedure that only relies on the model, which has been empirically identified. It results in an open loop linearization algorithm fully implementable in a digital architecture. The method allows exploiting the trade-off between the power-added efficiency (PAE) and the digital predistortion (DPD) computational load for a given linearity target.

A Fully Nonlinear Compact Modeling Approach for High-Frequency Noise in Large-Signal Operated Microwave Electron Devices

A technology-independent, inherently nonlinear approach is proposed for the compact modelling of high-frequency noise in microwave transistors under large-signal operating conditions. For the compact nonlinear noise model formulation we assume that noise generation can be described by a suitable set of distributed stochastic processes perturbing a very general, non-quasi-static deterministic description of the electron device, in accordance with the strategies adopted in physics-based methods for the choice and exploitation of microscopic diffusion noise sources. The propagation of the internal distributed noise sources up to the intrinsic device terminals leads to a set of non-stationary, correlated equivalent noise generators, nonlinearly controlled by the instantaneous large-signal working point of the device. Starting from a first formulation for the generators, formally derived from a physics-based description of the noise generation mechanisms widely adopted in distributed numerical modeling, mild approximations provide a fully behavioral representation that can be empirically extracted on the basis of measurement data only, and can be easily implemented into commercial computer-aided design tools by means of conventional, uncorrelated noise sources. As far as small-signal (i.e., linear) bias-dependent operation is concerned, it is shown how well-known, widely applied compact models for high-frequency noise can be considered as linearized special cases of the proposed approach. For a full validation, experimental examples are provided, both in small- and large-signal operation, for a GaAs-pHEMT, by considering the case study of a broad-band low-noise amplifier progressively driven into nonlinear regime by an increasing power interferer.