Reliability-Aware Support Vector Machine-Based High-Level Surrogate Model for Analog Circuits

Negative bias temperature instability (NBTI) has deleterious impact on threshold voltage and drive current of PMOS transistor. In this paper, a support vector machine (SVM)-based surrogate model (SM) for NBTI phenomenon is developed within the framework of the HSPICE MOSFET reliability analysis (MOSRA) model for gain and slew rate of a differential amplifier. Feasibility space identification and adaptive learning scheme are applied to improve the results using less number of training samples, which reduces run time. A single performance parameter evaluation of complete circuit requires 0.046 ms and 0.03 ms for gain and slew rate, respectively, which shows significant improvement over previous methodology. This SM is validated by finding a correlation coefficient between the SVM model and HSPICE. The values of the correlation coefficients are 0.9979 and 0.9997 for gain and slew rate, respectively.