Parasitic Effects Prediction in On-Chip-Antennas

In this article, an effective way to predict parasitic effects that arise with packaging is addressed with neural networks. A neural network model is trained with the dataset generated from electromagnetic simulations of a square spiral antenna operating in a sub-terahertz frequency range wrapped in a QFN package using Ansys High-Frequency Structure Simulator (HFSS). This helps to correlate the antenna design parameters and key metrics such as gain, radiation efficiency, and operating frequency, reflecting the parasitic effects of packaging. The proposed approach can be used as a reliable tool to understand the parasitic effects in OCAs, leading to faster design optimization of the antenna with enhanced overall performance and reliability.