Reliable leakage-enabled memristor model for large-scale circuits

Abstract

Memristors offer great potential for advanced memory and computing systems due to their ability to retain their resistance state. Several simulation models have been proposed to enable early analysis. However, there are convergence issues associated with some models, especially faster ones. This paper proposes reliable solutions to overcome convergence challenges in memristor simulation models. We studied and analyzed potential factors, including model nonlinearity, complexity, and the incorporated window functions. Adaptive solutions are developed to dynamically adjust to memristor behavior, effectively mitigating the convergence problem and improving accuracy and stability. We used genuine memristor experimental data and verified our solutions against the BELIEVER model in the simulations. These proposed adaptive techniques can enhance memristor convergence, enabling their adoption in diverse fields for improved simulation conditions. The maximum error of the proposed solution in the I–V characteristic remains below 15%. This level of accuracy is suitable, while it ensures the reliability of the circuit’s output with this specific model modification.