Neural Network-based Fast and Intelligent Signal Integrity Assessment Model for Emerging MWCNT Bundle On-Chip Interconnects in Integrated Circuit

At nanometer technology nodes, the efficient signal integrity and performance assessment of vast on-chip interconnects are crucial and challenging. For a long time, copper (Cu) has been used as an interconnect material in integrated circuits (ICs). However, as heading towards lower technology nodes, Cu is becoming inadequate to satisfy the requirements for high-speed applications due to its physical limitations. To mitigate this issue, a multiwall carbon nanotube bundle (MWCNTB) is proven to be a better replacement for Cu. Hence, the current work innovatively focuses on modeling, analysis, and performance evaluation of MWCNTB interconnects at 32 nm technology nodes using various machine learning (ML) and neural network (NN) based techniques for signal integrity assessment and fast computation of on-chip interconnect design. Based on the results obtained by comparing the different performance parameters, it is envisaged that NN-based ADAM technique leads to the best-suited model. The developed model is fruitful in evaluating the output performance of the system, such as power-delay-product (PDP), performing parametric analysis, and predicting optimum input design parameters of the driver-interconnect-load (DIL) system. This work utilizes HSPICE and Python electronic design automation tools for its implementation.