边距(机器学习)
电压
地铁列车时刻表
计算机科学
噪音(视频)
功率(物理)
灵敏度(控制系统)
噪声裕度
低压
MOSFET
电子工程
过程(计算)
晶体管
电气工程
工程类
人工智能
机器学习
物理
量子力学
图像(数学)
操作系统
标识
DOI:10.35848/1347-4065/acac3d
摘要
Abstract As a simple design method, the zoomed response surface (RS) method is useful for automatic design of low-voltage power MOSFETs. Low-voltage MOSFET characteristics have been improved continuously with respect to not only low power loss but also low cost in answer to the request for a high-performance system. Complicated requirements lead to a long development schedule and low yield. Model-based design and machine learning are prospective methods to address the problem. However, reported methods require high simulation numbers (>1000) for training to obtain high accuracy, and it is difficult to optimize parameters considering the process margin at the same time. This article shows a demonstration of a zoomed RS method for automatic design of 100 V-class power MOSFETs. Five parameters were automatically designed for not only minimizing on-resistance but also minimizing total power loss, taking into account process margin and switching noise with a simulation number of only 130.
科研通智能强力驱动
Strongly Powered by AbleSci AI