New mapping strategies for pre-optimized inductor sets in bottom-up RF IC sizing optimization

摘要

This paper presents new indexing and mutation operators, in the context of bottom-up hierarchical multi-objective optimization of radio frequency integrated circuits, for pre-optimized sets of solutions from the hierarchical sub-levels when moving up in hierarchy. Two ideas, one based on a Voronoi decomposition and another based on the nearest neighborhood, are explored, where, and unlike previous approaches that are based on sorting, the distance between elements determines the probability of decisions taken during optimization. Three implementations of those ideas were tried in AIDA's NSGAII evolutionary kernel, and successfully used in the optimization of a Voltage Controlled Oscillator and a Low Noise Amplifier with pre-optimized inductor sets obtained using the SIDeO toolbox, showing their strengths when compared to previous state-of-the-art mapping strategies.