Justin B. King

Nonlinear Electrothermal GaN HEMT Model Applied to High-Efficiency Power Amplifier Design

Gallium Nitride (GaN) high electron-mobility transistors (HEMTs) can operate at very high power-density levels, which may cause a significant temperature rise in the transistor channel. In addition, surface and substrate energy levels, or “traps,” can cause strong dispersion effects from pulsed I-V down to dc timescales. Such effects, for both simulation accuracy and device reliability purposes, must be accounted for in any nonlinear device model. In this paper, a novel nonlinear high-power GaN HEMT equivalent circuit electrothermal model is described. Features of the model include a nonlinear thermal subnetwork that is capable of capturing the well-known inherent nonlinear thermal resistance and capacitance of GaN material. Also included is a comprehensive dispersion model that can be extracted and modeled from simple measurements. The model can very accurately predict the pulsed I -V curves at different pulse widths and duty cycles from isothermal up to the safe-operating area limit. Large-signal one-tone, two-tone, and frequency sweep tests show excellent agreement with measurements. Finally, a continuous class-F amplifier is fabricated, and large-signal frequency sweeps are performed. Comparison between the measured and modeled amplifier metrics demonstrate that the model remains accurate over a 50% bandwidth under real-world conditions.

Padé-Approximation-Based Behavioral Modeling

A new, nonlinear, frequency-domain behavioral model, based on Pade approximation theory, is presented in this paper. The new Pade model is similar in concept to the X-parameter method, but it uses a rational function instead of the X-parameter models polynomial expression. The basic theory of the proposed Pade model and the model extraction methodology are provided. The proposed Pade approach increases the available modeling space beyond that available to the X-parameter method, and the performance of a selection of this family of models under various loading conditions is described in this paper. Along with the examples displaying the ability of the new model to predict fundamental frequency behavior, a second-harmonic example is also provided, with simulation results presented in both the frequency domain and the time domain. Based on the simulation studies and experimentally measured results, the new Pade model shows good accuracy and a high level of stability. All of these characteristics show that the proposed Pade model has high potential in RF design.

A wideband Sequential Power Amplifier

In this work, a wide bandwidth Sequential Power Amplifier (SPA) is demonstrated for the first time. Using wideband and high efficiency 10W and 45W amplifiers, an SPA is described, capable of delivering high efficiency at back-off without the bandwidth limitations of the traditional Doherty PA. Measurement results of the implemented amplifier are presented, with an average back-off drain efficiency of 59% achieved between 1.45 GHz and 2.4 GHz, and a peak back-off drain efficiency of 65% at 1.7 GHz.

A comprehensive electrothermal GaN HEMT model including nonlinear thermal effects

A novel nonlinear high-power Gallium Nitride (GaN) High Electron-Mobility Transistor (HEMT) equivalent circuit model is described. Features of the model include a nonlinear thermal subnetwork extracted using straightforward measurement techniques, and a modified Angelov/Chalmers single function drain current equation. The model can very accurately predict the Pulsed IV (PIV) curves at different pulse widths and duty cycles from isothermal up to the safe-operating area (SOA) limit, with high voltage drain-source pulses. Large-signal one-tone-test results are presented and show good fidelity with measurements for the first three harmonics, as well as accurate prediction of bias point shifting with increasing input power.

Equivalent circuit GaN HEMT model accounting for gate-lag and drain-lag transient effects

This paper introduces a full nonlinear compact equivalent circuit model (ECM) for a 10 W Gallium Nitride (GaN) high electron-mobility transistor (HEMT). The model accounts for low-frequency dispersion of both transconductance and output conductance by use of gate and drain filter networks applied to GaN for the first time. Other features of the model include an excellent fit of modelled drain-current to high-power pulsed-IV (P-IV) measurements up to 60 V across a range of quiescent points, using a single function drain-current equation. Global S-parameter fits across the bias plane have also been extracted. Large-signal harmonic sweeps are presented, showing a good fit to measured data over a range of bias points.

Nonlinear Behavioral Modeling Dependent on Load Reflection Coefficient Magnitude

A new frequency-domain nonlinear behavioral modeling technique is presented and validated in this paper. This technique extends existing Padé and poly-harmonic distortion models by including the load reflection magnitude, |ΓL|, as a parameter. Although a rigorous approach requires a full 2-D load-pull model to cover the entire Smith chart, simulation and experimental evidence have shown that such a 1-D model-that retains only amplitude information of the load reflection coefficient-can give accuracy close to that of a full 2-D load-pull model. Consequently, neglecting the phase constitutes an approximation that provides large benefits without appearing to lead to a severe compromise in accuracy. Furthermore, compared with traditional load-independent models, the new |ΓL|-dependent models provide a major improvement in model accuracy. After a discussion of the model extraction methodology, examples are provided comparing traditional load-pull X-parameter models with the model presented in this paper. The new model not only provides consistently good accuracy, but also has a much smaller model file size. Along with the examples that display the ability of the new modeling technique to predict fundamental frequency behavioral, a second harmonic example is also provided. The modeling approach is also validated using measurements results.

A novel continuous Class-F mode Power Amplifier

In this work, a novel, higher-order continuous Class-F mode Power Amplifier (PA) is presented for the first time. The described continuous mode expands the design space available for the development of wide-bandwidth and high-efficiency PAs. The current and voltage waveforms are derived for the mode of operation, along with the impedances to be presentated at the current generator plane. Measurement results of a 10W PA, operating from 1.1GHz to 2.6 GHz, are presented to validate this new mode.

An improved quadratic poly-harmonic distortion behavioral model

In this paper, the basic quadratic form of the poly-harmonic distortion model is first presented and this is then extended to provide a new, modified quadratic poly-harmonic distortion model. Comparisons between the X-parameter model, the basic quadratic poly-harmonic distortion model, and the modified version are provided. Both simulation and experimental test results show that the new modified model provides significant improvements in accuracy, not only for the fundamental frequency, but also for DC. Work on the optimization of the model is also presented, providing further improvements in both the model extraction time and the file size.

The Continuous Harmonic-Tuned Power Amplifier

In this work, a new power amplifier (PA) mode is introduced-the continuous Harmonic-Tuned 2nd (HT2) mode. The mode offers the potential to deliver high efficiency across wide bandwidths. The voltage and current waveforms required to realise this novel mode are examined, theoretical performance is described, and measurement results are presented, verifying the theory.

A new nonlinear behavioral modeling technique for RF power transistors based on Bayesian inference

A novel nonlinear behavioral modeling technique, for transistor behavioral modeling, is presented in this paper. Compared with existing modeling techniques, the new approach is based on a fundamentally different theory, Bayesian inference (one of the core methods of machine learning). The new technique not only good at handling multidimensional modeling problem, it could also greatly alleviated the notorious overfitting issue through corresponded model extraction method. Both simulation and experimental test examples for a 10W Cree GaN transistor are provided. The new model provides accurate prediction throughout the Smith chart at different input power levels.