Evolving sinusoidal oscillators using genetic algorithms

In the present paper, single-opamp sinusoidal oscillators are synthesized using genetic algorithms. The motivation is to evolve new topologies of oscillators using different active building blocks (ABBs) and automate the study of their properties. A new fitness evaluation scheme by analyzing transfer function of the circuits is used and a learning scheme loosely inspired from Lamarckian search is also suggested. A new problem specific crossover operator is tested and a comparative study of different crossover operators is done. On comparison of the results of the GA with existing results, it was found that the GA rediscovered all the twelve canonic singleopamp based SFOs. Some new interesting opamp, OTRA and DDCC based topologies of oscillators are also presented. It is clearly explained how this study can be extended to other ABBs or multiple ABBs.