A scalable model for on-chip spiral inductors by Space-mapping neural network

This paper presents a scalable model for on-chip spiral inductors utilizing the space-mapping neural network (SMNN), built upon an enhanced 1-π equivalent circuit model. To address the inductor drop observed at low frequencies, a series network consisting of two pairs of inductors and resistors is integrated into the 1-π equivalent circuit model. The topology of the SMNN is optimized using the global search capability of the particle swarm optimization (PSO) algorithm, which determines the optimal weights and thresholds for the SMNN. Compared with traditional neural network models, the proposed SMNN reduces the number of neurons required for mapping the input-output nonlinear relationship, leading to enhanced accuracy and reduced modeling time, which takes a mere 10 to 15 seconds.