A Parameter Extraction Method for LC Circuit of DB-BPF Based on Fully Connected Network

A parameter extraction method for the LC circuit of a dual-band bandpass filter (DB-BPF) based on the fully connected network is proposed in this article. The network learning process consists of two stages: 1) pretraining and 2) fine-tuning. The network after pretraining has the ability to output the inaccurate values of LC circuit parameters with the desired ${S}$ -parameters (obtained according to the filter design requirements) as input. The precise LC circuit parameters can be obtained from the network during the subsequent fine-tuning process. This parameter extraction method can be applied to LC circuits of various bandpass filters. In addition to the amplitude of S-parameter, the phase of S-parameter is also used as learning data for the pretraining network. Comparative experiments show that the most accurate LC circuit parameters can be predicted using the network trained with phase information compared with the network trained without phase information. Two 7-order narrowband and two 9-order wideband DB-BPFs with four transmission zeros at different center frequencies are designed to demonstrate the validity of the proposed method, and the experimental results show that the proposed method is valid and effective.