Frank Chau

The Safe Operating Area of GaAs-Based Heterojunction Bipolar Transistors

The safe operating area (SOA) of GaAs-based heterojunction bipolar transistors has been studied considering both the self-heating effect and the breakdown effect. The Kirk effect induced breakdown (KIB) was considered to account for the decrease of the breakdown voltage at high currents. With reasonable emitter ballastors, the KIB effect was shown to be the major cause for device failure at high currents, while the thermal effect controls the low current failure. The effect of emitter resistance and base resistance on device stability was also studied. While the emitter resistance always improves the device stability by expanding the SOAs, the base resistance degrades SOAs when the KIB dominates the failure mechanism. The effect of the base resistance on SOAs was explained by its control on the flow of the avalanche current. Since the KIB effect depends on the collector structure, it was shown that a nonuniformly doped collector can effectively improve the SOAs

Reliability Study of InGaP/GaAs HBT for 28V Operation

This paper reports the device process and reliability aspects of a high voltage InGaP/GaAs HBT technology for 28V operation. The key differences and challenges from the material and process perspectives relative to the conventional low voltage HBT technology are first described. Long-term reliability tests performed at 28V bias voltage, 5.2kA/cm2 current density and 310degC junction temperature resulted in zero device failures after over 3600 hours of stress. Wafer-level reliability tests using extreme current stress conditions were used to force transistors to degrade in short time for quick checking of transistor integrity and process reliability. Transistor lifetimes were generally higher than those in conventional low voltage HBTs at similar junction temperatures. Combining with the previously reported ruggedness and linearity performance, the high voltage InGaP/GaAs HBT is a mature technology for use in the 28V high linearity power amplification

A scalable high power nonlinear HBT model for a 28V HVHBT

A scalable nonlinear HBT model for a Building Block (1BB) of 28V InGaP/GaAs HBT is presented. It is based on the AgilentHBT (AHBT) model. The building block consists of 32 finger HBTs, an input pre-matching circuit and a transistor used as an emitter follower in the related bias circuit. The P1dB of 1BB is 32.5dBm. The model can account not only for DC, thermal, junction capacitances, S-parameters but also RF power, gain, IM3, operation current and collector efficiency. A good match between simulation and measurement has been achieved. By using a multiplicity parameter the model can accurately predict the DC and nonlinear RF performance of two Building Blocks (64 fingers, P1dB of 35.7dBm) and four Building Blocks (128 fingers, P1dB of 38.2dBm).

An accurate packaged model for HVHBT 120W power amplifiers and its application to 250W Doherty amplifiers

An accurate yet straightforward modeling approach is reported to develop a packaged model for InGaP/GaAs HVHBT 120W power amplifiers (PA) at 2140MHz. The model consists of an accurate HBT model, models for internal pre-match circuits and models for coupled bond-wire arrays. The HBT model is scaled up from an AHBT (Agilent HBT) model for a unit cell transistor. The models for the internal pre-match circuits are lumped-element equivalent circuit models derived from the EM simulated S-parameters. For high power amplifiers, the coupling between the bond-wires must be accounted for. In this work, the coupling between the bond-wires is determined up to the 4th next bond-wires. The model of 120W PA can precisely simulate small signal S-parameters as well as large signal performances of harmonic load-pull power sweep, such as output power, P1dB, G1dB, efficiency and operation current. A very good agreement between simulation and measurement has been achieved. The model is then applied to simulate a 250W symmetric Doherty amplifier with an efficiency above 70%. The results of simulation agrees well with the measurement. The difference in efficiency between simulation and measurement is only 1.6%.