Automatic parametric model development technique for RFIC inductors with large modeling space

We present an automatic method to extract parametric model for RFIC inductors in large modeling space covering a wide range of geometrical variables. We use a modified double-pi network as the equivalent circuit topology of the inductor model. Lumped element values are computed using empirical functions which are formulated in terms of inductor geometries and numerical coefficients. The automated method extracts coefficients through optimization of circuit model and electromagnetic (EM) data. An intelligent mapping scheme is formulated to map geometries of inductors to equivalent circuit components using neural networks making the model suitable for handling wide range of geometrical variations. Model developed in this way shows good accuracy compared to EM data with a significant reduction of developmental cost.