An ultra-fast and precise automatic design framework for predicting and constructing high-performance shallow-trench-isolation LDMOS devices

Abstract

The shallow trench isolation-based laterally diffused metal–oxide–semiconductor (STI LDMOS) is a crucial device for power integrated circuits. In this article, a novel framework that integrates an optimal objective function, Bayesian optimization (BO) algorithm, and deep neural network (DNN) model is proposed to fully realize the automatic and optimal design of STI LDMOS devices. On the one hand, given the structure of the device, the DNN model in the proposed method can provide ultra-fast and highly accurate performance estimation including breakdown voltage (BV) and specific on-resistance (R_onsp). The experimental results demonstrate 98.68% average prediction accuracy for both BV and R_onsp, higher than that for other machine learning (ML) algorithms. On the other hand, to target the specified value of BV and R_onsp, the proposed framework can fully automatically and optimally design the precise device structure that simultaneously achieves the target performance with the optimal figure of merit (FOM) of the device. Compared to technology computer-aided design (TCAD), there is only a 0.002% error in FOM and a 2.83% average error in BV and R_onsp. Moreover, considering the training time of the DNN model, the proposed framework is 100 times as efficient as other conventional frameworks. Thus, this research provides the experimental groundwork for constructing an automatic design framework for an LDMOS device and opens new opportunities for accelerating the development of LDMOS technology in the future.