Analytic and SPICE modeling of stochastic ReRAM circuits

TL;DR

This paper reviews and improves SPICE-based modeling of stochastic ReRAM circuits, addressing computational challenges and identifying attractor states in memristive systems.

Contribution

It introduces an enhanced SPICE implementation of the master equation approach for stochastic ReRAM circuits and discovers attractor states in driven memristive systems.

Findings

Improved SPICE implementation of stochastic circuit modeling.

Identification of attractor states in memristive circuits.

Addressed computational challenges in analyzing stochastic ReRAM circuits.

Abstract

The modeling of conventional (deterministic) electronic circuits - ones consisting of transistors, resistors, capacitors, inductors, and other traditional electronic components - is a well-established subject. The cycle-to-cycle variability of emerging electronic devices, in particular, certain ReRAM cells, has led to the concept of stochastic circuits. Unfortunately, even in relatively simple cases, the direct transient analysis of stochastic circuits is computationally demanding and potentially impractical, if possible at all. An important development in this area has been the application of a master equation that is easily implemented in SPICE. In this conference paper, we briefly review the master equation approach and present an improved implementation of this approach in SPICE. Moreover, we find an attractor state in a periodically driven memristive circuit - a stochastic counterpart of deterministic memristor attractors.