An efficient ternary multiplier for enhanced on-chip AI wearable systems using graphene nanoribbon field-effect transistors

Abstract

Enhanced on-chip AI wearable systems are transforming the interaction between individuals and technology, facilitating the development of smarter and more responsive devices that integrate seamlessly into everyday life. As the demand for such advanced systems increases, the necessity for energy-efficient circuits that provide extended battery life alongside high-performance processing becomes paramount. Within this framework, the graphene nanoribbon field-effect transistor (GNRFET) presents a promising solution, particularly due to its capability for ternary processing, which significantly enhances response speed, reduces power dissipation, and improves area efficiency.This paper presents an innovative ternary multiplier (TMUL) utilizing GNRFET technology, specifically designed to achieve reduced propagation delay, lower power consumption, and increased robustness against process variations, all while minimizing the transistor count. The proposed architecture incorporates a single transistor within the logic pathways leading to the outputs, integrates a greater number of high-threshold voltage (V_th) devices, and employs unary operators. These characteristics collectively advance the design objectives.The efficiency of the proposed TMUL is evaluated in comparison to seven contemporary TMUL circuits, employing a 32-nm GNRFET technology node with a supply voltage of 0.9 V. Remarkably, our design exhibits improvements of at least 8.98 % in delay, 2.22 % in power consumption, and 40.88 % in energy efficiency. Further performance assessment through Monte Carlo simulations confirms the superiority of the proposed design, underscoring its enhanced reliability amidst process variations and its minimal variance in energy consumption. By integrating the proposed single-trit TMUL to 3-trit TMUL, the projected circuit requires 1139 GNRFETs, with low power consumption and high speed operation.