An Automatic ReRAM SPICE Model Generation Methodology for ReRAM-Circuit Co-Simulation
서재현
Abstract
서재현
Abstract
This paper presents an automatic ReRAM SPICE model generation methodology which enables true ReRAM-circuit co-simulation in standard SPICE. In our method, a model generation tool automatically produces SPICE models of various ReRAM devices and selectors from the measured I-V data including newly measured ReRAM's SET behavior in our proposed experiment. The generated models are compatible with standard SPICE and can describe diverse ReRAM behaviors including nonlinear I-V characteristic, I-V relationship during SET, and dependency of I-V curve on SET current. Because describing these behaviors in standard SPICE has been the critical obstacle to simulate ReRAMs and circuits together, especially in multi-level cell simulation, our method can be a key enabler of ReRAM-circuit co-simulation. To verify our method, SPICE models for diverse ReRAMs were generated and simulated. The results show that our model can accurately describe the original data. To demonstrate the usefulness of our method, we also simulated and analyzed example ReRAM circuits such as 1R and 1S1R arrays and a complex CMOS sensing circuit for ReRAM. These simulations and analyses show that our method enables quantitative ReRAM analyses such as trade-off analyses for reading margin and sensing delay.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This paper presents an automatic ReRAM SPICE model generation methodology which enables true ReRAM-circuit co-simulation in standard SPICE. In our method, a model generation tool automatically produces SPICE models of various ReRAM devices and selectors from the measured I-V data including newly measured ReRAM's SET behavior in our proposed experiment. The generated models are compatible with standard SPICE and can describe diverse ReRAM behaviors including nonlinear I-V characteristic, I-V relationship during SET, and dependency of I-V curve on SET current. Because describing these behaviors in standard SPICE has been the critical obstacle to simulate ReRAMs and circuits together, especially in multi-level cell simulation, our method can be a key enabler of ReRAM-circuit co-simulation. To verify our method, SPICE models for diverse ReRAMs were generated and simulated. The results show that our model can accurately describe the original data. To demonstrate the usefulness of our method, we also simulated and analyzed example ReRAM circuits such as 1R and 1S1R arrays and a complex CMOS sensing circuit for ReRAM. These simulations and analyses show that our method enables quantitative ReRAM analyses such as trade-off analyses for reading margin and sensing delay.
Key concepts: Spice, Resistive random-access memory, Electronic engineering, Computer science, Set (abstract data type), CMOS, Memristor, Electronic circuit