An Efficient Fixture Removal Embedded Modeling Method Based on TDR and CNN Technique

S-parameters are typically employed in equivalent circuit (EC) modeling of electronic devices. However, for existing modeling procedures, the impact of fixtures on S-parameter measurement cannot be neglected and needs to be eliminated through de-embedding before modeling. This letter proposes a new approach that integrates fixture removal into modeling procedure by combining time-domain reflection (TDR) and convolutional neural network (CNN) techniques. The proposed approach bypasses the need for separate de-embedding, allowing for direct derivation of the EC model. Different to the traditional modeling procedure, its advantages in simplifying the modeling procedure and avoiding the errors introduced by de-embedding have been validated by the physical measurement.