A Scalable Model of Board to FPC Interconnect Using Neural Networks

In this paper, we propose an equivalent circuit model which includes the board to flexible printed circuit (FPC) connector transition region and the FPC connector to FPC cable transition region. Full wave simulations of the FPC interfaces are performed and the s-parameters of the simulators are compared to those of the vector network analyzer (VNA) measurements. The simulation shows very similar s-parameters compared to the VNA measurements, confirming its usefulness as a reference tool for the equivalent circuit model. Lumped capacitors and inductors are used as the parameters of the equivalent circuit model. We extract the parameters by fitting the s-parameters of the equivalent circuit model to those of the 3D full wave simulator. The frequency range of the proposed model is 500 MHz to 3 GHz which is suitable for use with modern high speed applications. The neural network approach is then introduced as a novel behavioral modeling method for the board to FPC interface. The length of the board to FPC interface is considered as a scalable parameter and used as the training data of the neural network structures. The predictions of the s-parameters using the neural networks, when an arbitrary transition length is given, are very good compared to the 3D full wave simulations, and these results validate the neural network as a promising candidate modeling technique for modern high-speed interconnects.