Optimization of physics-based equivalent circuits for microstrip patch antennas

While many commercial EDA tools already have the capability to synthesize networks by evolutionary and statistical algorithms, we gain a lot of physical insight when the network synthesis problem is solved by gradient-based local optimization instead. Given an estimation of the component values of the equivalent network with physics-based or empirical formulas for a microstrip patch antenna, the cost function to be optimized is close to convex near the initial guess. Moreover, the disparate sensitivities of the cost function with respect to optimization variables depends on the resonant frequency of the patch antenna and can be equalized by frequency scaling. The paper demonstrates how physics-based equivalent circuits can be automatically corrected to facilitate patch antenna design.