Dielectric Substrate Prediction Through Transmission Measurements and Machine Learning

Dielectric properties of the substrates play an important role in the design and performance characterization of communication components such as antennas, filters, and sensors. Conventionally, dielectric probes are used to measure the properties of the substrate. However, the dielectric probes are very expensive and easily breakable instruments. In this work, a novel method of dielectric substrate prediction has been proposed using S 12 measurements with two waveguides, along with the application of machine learning. Extensive data collection is done using multiple simulations of the proposed method in 3D electromagnetic software in the X‐band frequency range (8–12 GHz). The measurements are then conducted by using two waveguides, and the data is compared with the simulation data set, where the decision is made based on the comparison of dielectric properties. For verification of the proposed method, dielectric substrates of FR4 and Rogers 5880 have been used, which demonstrated very close agreement between the measured properties and properties from the data sheet.