Modeling and Analysis of a Large-Scale Memristor Crossbar Array for Neuromorphic Chip

Abstract

This paper presents a complete equivalent circuit model of a large memristor crossbar array, including parasitic capacitance, inductance, and resistance. By changing the rise time of the input signal and monitoring the output voltage of different output ports, signal integrity (SI) problems such as crosstalk, and IR voltage drop in the memristor crossbar array are analyzed. The simulation results show that with the decrease of the rise time, the electromagnetic effect will become obvious, which will lead to the crosstalk ripple and distortion of the output signal becoming serious. And as the output port moves away from the input, the IR voltage drop increases with the increase of the coupling elements. This will point the direction for improving the performance of memristor crossbar arrays. Furthermore, different from conventional signals, we propose to simulate spiking signals using the Izhikevich neuron model for more accurate analysis.