An enhanced multi-objective evolutionary algorithm (MOEA/D-DE) for the applications of analog sizing with both W/L and a novel operating point driven (OPD) based methods

Abstract

In today’s electronics world, due to the growing requirements of mixed signal VLSI designs and the SoCs (system on chip), the design complexity is increasing drastically. Since the well-designed CAD tools can easily support the design of the digital circuits, analog CAD tools are still not enough for the needs of mixed signal VLSI designs and SoCs. One of the biggest reasons for this deficiency is the complex design procedure of an analog system. To design an analog circuit is much harder than designing a digital circuit. However, the world we live in is analog and there is no way to avoid analog circuitry. For all these reasons strong algorithms are trying to be implemented to automate analog circuit design. The sizing problem of analog circuits to obtain the best performance is an important subject of analog design automation. First of all, the reason why Evolutionary Algorithms are used for the analog sizing problem has been explained. Then, a Multiobjective Evolutionary Algorithm based on the Decomposition of the objective functions has been used as a background work. Lots of improvements have been realized to improve the quality of this method for more complex analog circuit sizing problems. Also, the optimization variables have been changed to DC operating points of the transistors instead of W/L values, in order to improve the search space. All the methods were implemented and the results are given in the work. It can be seen that the proposed W/L or the novel OPD (operating point driven) based methods are so powerful algorithms to optimize the analog sizing problem. During the thesis work, a folded cascode amplifier and a gain boosted amplifier have been optimized..