Taguchi Approach and Response Surface Analysis for Design of a High-performance Single-walled Carbon Nanotube Bundle Interconnects in a Full Adder
In this study, it was attempted to design a high-performance single-walled carbon nanotube (SWCNT) bundle interconnects in a full adder. For this purpose, the circuit performance was investigated using simulation in HSPICE software and considering the technology of 32-nm. Next, the effects of geometric parameters including the diameter of a nanotube, distance between nanotubes in a bundle, and width and length of the bundle were analyzed on the performance of SWCNT bundle interconnects in a full adder using Taguchi approach (TA). The results of Taguchi sensitivity analysis (TSA) showed that the bundle length is the most effective parameter on the circuit performance (about 51% on the power dissipation and 47% on the propagation delay). Moreover, the distance between nanotubes greatly affects the response compared to other parameters. Also, response surface method (RSM) indicated that an increase in the length of interconnects (L) improves the output of power dissipation. As the width of interconnects (W) and diameter of CNTs (D) increase the power dissipation also increases. Decrease in the distance between CNTs in a bundle (d) leads to an increase in power dissipation. The highest value of power dissipation is achieved if the maximum values for the parameters of length and width of interconnects (L, W), and diameter of CNTs (D) and the minimum value of the distance between CNTs in a bundle (d) are considered. It is also revealed that an increase in the length of interconnects (L) increases the propagation delay. Eventually, the optimum parameters are reported and the performance of the optimized system is compared using different methods (TA and RSM). Results indicate that the difference between the performance of optimal design of SWCNT bundle interconnects in a full adder predicted by different methods is less than 6% which is acceptable according to engineering standards.