Nanoscale CMOS Modeling based on Artificial Neural Networks: Design of SRAM Circuit for Reliability

In this paper, we applied the artificial neural networks (ANN) algorithm to modeling techniques to accurately and quickly predict the characteristics of the nanoscale CMOS. Since ANN only relies on data to predict results, it is a very strong point that semiconductor modeling may take into account less modifications by existing physical theories. Then we implemented the model in the Verilog-A language. It was confirmed that the results were suitable with high accuracy in both the model stage and the circuit stage. In addition, our reliability approach is verified through the SNM analysis of SRAM circuits. Therefore, in simulator environments such as SPICE, our proposed the ANN model is suitable for application to various nanoscale circuits.