Modeling of power supply noise with non-linear drivers using recurrent neural network (RNN) models

Recurrent Neural Network (RNN) functions can be used to model highly nonlinear driver circuits. In this paper, RNN modeling technique is extended to multiple ports to model both the power supply noise and the ground noise accurately. RNN driver models can be extended to multiple ports to capture sensitive effects like Simultaneous Switching Noise (SSN) accurately when multiple driver circuits are switching. A comparison study is performed on test cases between RNN models and transistor level driver circuits for accuracy and computational speed-up is performed on few test cases. Results show that RNN driver models have huge computational speed-up over transistor level driver circuits.