A New Optimizer for Performance Optimization of Analog Integrated Circuits

In this paper, a new optimizer for multiple criteria constrained performance optimization of analog integrated circuits, based on non-linear programming and heuristic techniques is presented. Biggs´ dominant degree method, which effectively handles nonquadratic functions is used for gradient based optimization. A modified Parkinson Hutchinson Simplex algorithm and Guided Random Search technique are used to perform optimization based only on function evaluations. These optimization techniques are combined effectively to overcome numerical difficulties encountered during circuit simulation and gradient based optimization. Experimental results obtained using the new optimizer on real life circuits at Texas Instruments Inc. show consistent improvements in circuit performance.