Analysis of a New Evolutionary System Elitism for Improving the Optimization of a CMOS OTA

Analog integrated circuits design is a complex task due to the large number of input variables that must be determined in order to achieve different design goals such as voltage gain, unit voltage gain frequency, phase margin and dissipated power. This paper describes and implements an evolutionary optimization solution based on genetic algorithms and the well-known SPICE simulator, named "AGSPICE/FEI", with current industry standard BSIM3v3 model that has the capability of searching solutions that better comply with the multiple design goals specified a priori of a single-end, single-stage operational transconductance amplifier (OTA). The AGSPICE/FEI can provide a large set of solutions allowing the designer to choose the best solutions, which fully meet the design goals, and, thus, analyze the performance obtained by the modified genetic algorithm, which performs two types of elitism: one conventional and another non-conventional. This paper performs the study of the performance of the non-conventional elitism by comparing the performance achieved by several different groups, in which in one of them is applied only the conventional elitism and in the other groups are applied the non-conventional elitism in different settings. The results have been analyzed for different design specifications. The experimental results have demonstrated that this new elitism stage is able to increase the speed of the AGSPICE/FEI searching process and also might provide solutions that better suit the design goals.