Radial Basis Function Based Surrogate-Assisted Metaheuristic Approach for Variability Analysis

Variability analysis of analog circuits is an impor-tant aspect of integrated circuit design, aiming to address the inherent uncertainties and variations in manufacturing processes. Analog circuits are particularly susceptible to process variations, as even small changes in design parameters can significantly impact their performance, affecting key metrics like bandwidth, gain, linearity, and power consumption. This work presents a novel approach for efficient and fast variability analysis using radial basis function based surrogate-assisted (SuA) metaheuris-tic algorithms. The proposed approach is benchmarked with existing methods of variability analysis available in the literature by conducting two case studies: the first involves estimating the variability in phase noise for RF CMOS LC tank oscillator, and the second concerns the variability in the amplitude of differential output signal in current mode LVDS driver. A significant reduction in run time is reported.