BTI-Aware Cell Characterization based on Neural Network

초록

As semiconductor technology scales down, Bias Temperature instability (BTI) became more important in the presence of reliability of CMOS circuits. Also scaling down of the circuits has more effect on the performance due to the aging effect, leading to an increase in time complexity for cell characterization. In this paper, to cope with the time complexity issues, we propose neural network based cell characterization considering BTI to predict gate delay faster and accurately. Our proposed method extracts the delay time of gates with HSPICE simulation to use as training data for the neural network. The model generated after training with the neural network can predict the gate delay of specific conditions. Experiment results show that our model can reduce cell characterization time 700x faster compared to HSPICE with 98.15% accuracy.

키워드