Home Featured Articles Rapid Model-Based Evaluation Board Success for a 160W L-Band GaN PA 

Rapid Model-Based Evaluation Board Success for a 160W L-Band GaN PA 

642
0

by Rached Hajji and Kim Tran, Qorvo USA, Inc. & Larry Dunleavy and Laura Levesque, Modelithics, Inc.

The demand for Power Amplifiers (PAs) continues to grow with the increase in industrial and military applications of wireless and microwave technologies.  Current PA designs must meet increasingly challenging performance goals in terms of efficiency, gain, linearity, power and bandwidth. The optimal design of the complete PA usually requires some compromise between several of these goals, and the ability to closely analyze the circuit under various conditions helps engineers make these design decisions. With sufficiently accurate nonlinear models, these trade-offs can be optimized in simulation prior to building any hardware. This is true for evaluation board (EVB) reference designs as well as final application circuits.

In fact, simulation-based PA design with non-linear models is the best path to meeting these complex design challenges. An alternative is to use measured load-pull data to determine matching condition trade-offs for PA design. However, load-pull characterization is time-consuming, requires specialized equipment, is not always available to designers, and its use is limited to the specific biases, frequencies and impedance ranges used to take the data.

Figure 1: Model information datasheet for the TGF2819-FS non-linear GaN HEMT model, including outline of model features, many model-to-measurement performance plots, and clear definition of reference planes

Access to compact non-linear transistor models greatly simplifies PA design and enables extrapolated load-pull and other non-linear simulations that can extend beyond the bench conditions used to take the data used to build the model. Simulated-to-measured agreement can be increased further by accurately accounting for parasitics and specific characteristics of the passive components used in the input and output matching networks during the electronic design stage. In this work, a 160W GaN Class AB power amplifier EVB example is described whose design was based solely on simulation-based design optimization in Keysight Technologies Advanced Design System (ADS), with advanced models. The following summary will illustrate the design process and show how successful first-pass design is possible with simulation, even for a prototype with challenging performance goals.

Power Amplifier Design Example Overview

The design example to be presented is a power amplifier EVB prototype that can deliver 52dBm (160W) P3dB minimum with at least 50% power added efficiency (PAE) over the frequency band 1.35 GHz to 1.75 GHz. The selected transistor device is Qorvo’s TGF2819-FS packaged GaN on SiC HEMT RF transistor. The device is based on Qorvo’s GaN25HV technology and features an operating frequency range of DC to 4 GHz and rated output P3dB power of 54dBm. A non-linear model for this device is contained within the Modelithics-Qorvo GaN Library, described elsewhere1,2.

The described EVB development relies solely on simulation-based design using Modelithics’ non-linear model for the Qorvo TGF2819-FS device, a temperature and bias-dependent non-linear model validated with single tone power and load-pull at 2 GHz and 3 GHz. So, there is no load-pull data available in the desired design band, but the model was validated for broadband S-parameter fitting over 0.1 to 8 GHz. The GaN HEMT non-linear model features and performance plots are detailed in a model data sheet (Figure 1).  For designer convenience and reference, similar model data sheets are available for all Modelithics active and passive models.  The non-linear simulations and optimizations of the EVB were done using harmonic balance analyses in Keysight ADS and the matching networks were optimized using Momentum Electromagnetic (EM) co-simulation with Modelithics’ scalable Microwave Global Models™.

Design: Determine Source and Load Impedances

The first step of the PA design is often to determine target source impedances across the operating band. A small-signal simulation and source-pull of the discrete GaN HEMT model was done at the package reference planes. A small-signal source-impedance match design using ideal matching network elements was followed by an HB simulation, then an EM co-simulation of the lumped/distributed input matching network, as outlined in Figure 2.

Figure 2: Design steps for optimum source impedance determination and input matching network design. Bias = 50V, 250mA
Figure 3: Design steps for optimum load impedance determination and output matching network design. Bias = 50V, 250mA

Next, the optimum load impedances were determined using a large signal load-pull simulation across the operating band, optimized for both max power and max efficiency.

The load impedance chosen to design the output matching network required some compromise between the maximum power and maximum efficiency simulated impedances. The resulting output matching network was evaluated with EM co-simulation and Modelithics capacitor and inductor models, and was compared to the target impedances selected from non-linear transistor model load pull simulation, as the plot of Figure 3 shows.

Further analysis of the output matching network design was performed to check the 2nd harmonic impact on power and efficiency across the operating band. This is to verify that the PA performance is not degraded inadvertently by a poor second harmonic condition, and is particularly important when the power discrete device does not have output pre-match inside the package (see Figure 4).

Build and Validate PA Evaluation Board

The power amplifier design was finalized and fabricated based completely on the predicted performance of the simulations using the non-linear GaN HEMT model and the parasitic capacitor and inductor models. The assembled prototype EVB and complete simulation schematic are shown in Figure 5.

Figure 4: Second harmonic simulation to verify acceptable PA performance at 2fo load impedance. Bias = 50V, 250mA
Figure 5: Assembled EVB prototype fixture (left) and complete power amplifier EM co-sim evaluation circuit in ADS (right)
Figure 6: Wideband small signal S-parameter results of the PA EVB design, including accurate harmonic response prediction. Bias = 50V, 250mA. Red solid lines = model data, Blue dashed lines = measured data

The measured performance of the EVB prototype had excellent overall model-to-measurement agreement in essentially every aspect across the full PA operating frequency band. The wideband small signal predicted performance of the simulation aligned very well with the measurement, including the prediction of response in the second harmonic frequency band (Figure 6). The large signal harmonic balance power sweep simulations showed accurate prediction, exceeding minimum design goals throughout the PA band, with more than 70% PAE, >52 dBm P3dB, and 3dB gain (G3dB) of 14-15dB (Figures 7 and 8). A summary of the PA design goals, simulation and measurements is outlined in Table 1.

Table 1: Initial PA design example goals compared to simulation-based performance prediction and actual measured PA performance with no bench tuning

Summary 

A 160W GaN PA evaluation board example illustrates the excellent results that are possible using a simulation-based design flow with high accuracy models.  The use of non-linear models, such as those in the Modelithics-Qorvo GaN Library, enables broad-band complex and multi-harmonic predictions and design optimizations. Design cycles for new EVBs and application-specific PA circuits are reduced significantly by eliminating the need for custom device characterization in cases where available load-pull data does not cover design application space. Design time is also reduced by enabling first pass design success so that extensive bench tuning is not required, or is minimized. This use of advanced scalable parasitic simulation models and EM co-simulation for matching and bias elements is also an important aspect of implementing efficient and successful PA design flows.

Figure 7: Large signal power sweep gain (dB) and drain efficiency (%) at 1.35 GHz (left), 1.55 GHz (middle) and 1.75 GHz (right). Bias = 50V, 250mA. Red solid lines = model data, Blue dashed lines = measured data
Figure 8: Large signal 3dB compression simulation results over the PA operating frequency band 1.35 GHz to 1.75 GHz. P3dB (left), G3dB (middle) and drain efficiency (right). Bias = 50V, 250mA. Red solid lines = model data, Dashed lines = measured data for two assembled units

Acknowledgements and Additional Information 

The simulation-based design process and results were presented as part of an exhibitor workshop presented by Modelithics, Inc., Qorvo USA and Keysight Technologies at the IMS 2017 conference held in Honolulu, HI in June 2017. The authors would like to thank Jack Sifri of Keysight Technologies    for collaboration related to this work and his review of the manuscript.

Free access to the Modelithics® Qorvo GaN Library is available to approved designers, and can be requested at:  https://www.modelithics.com/mvp/qorvo. For designer convenience, an extensive collection of documentation and example workspaces related to the Modelithics Qorvo GaN device models is included in the library installation.

References 

1L. Dunleavy, H. Morales, C. Suckling and Kim Tran, “Device and PA Circuit Level Validations of a High Power GaN Model Library,” Microwave Journal, August 2016.

2L. Dunleavy, J. Liu, M. Calvo, H. Morales, L. Levesque and R. Santhakumar, Advanced Nonlinear and Noise Modeling of High Frequency GaN Devices,” Microwaves & RF, Nov. 2017.

(642)

print

LEAVE YOUR COMMENT