基于4H-SiC MOSFET的ANN建模用于极端温度应用的SiC放大器设计与验证

This study presents a high-precision artificial neural network (ANN)-based modeling framework for SiC MOSFETs, achieving high accuracy with less than 1.2% error over a wide temperature range (27 °C–500 °C). The developed model enables reliable SPICE simulations of SiC integrated circuits, facilitating the design and experimental verification of high-temperature analog circuits. A single-stage common-source (CS) amplifier with a pMOS current source load demonstrates a maximum low-frequency gain of 25.5 dB, while a two-stage amplifier achieves 52.5-dB gain with a 220-kHz unity-gain bandwidth (UGBW) at 500 °C. Experimental results confirm the agreement between simulation and measurement, with deviations of less than 6.9% in the 1-MHz frequency domain for key performance indicators at both room temperature and 500 °C operating conditions. This study establishes a robust and scalable methodology for the design and verification of SiC-based circuits over a wide temperature range, with potential applications to other emerging semiconductor technologies operating in extreme environments.