Estimation of Radiated Emissions from microstrip PCB using neural network model

Printed Circuit Board (PCB)-traces are one of the sources of PCB Radiated Emissions (RE). Due to the ever-increasing clock speeds on PCB, these traces have become electrically long resulting in higher RE and consequently, failure to comply with the limits imposed for electromagnetic compatibility (EMC) compliance. Therefore, it is crucially important to estimate these emissions at the design phase of the electronic device to ensure time and cost savings for putting in marketplace. In this paper, a simple Multi-Layer Perceptron (MLP) model has been developed for estimating electromagnetic emissions of PCB-trace. This proposed model was trained, tested and validated using a large amount of data sets collected from HFSS simulation. Various PCBs were employed to verify this proposed model using both HFSS simulation and measurement in Semi-Anechoic Chamber (SAC). The results from the proposed model were then compared to the results obtained from 3-D HFSS full-wave simulation and the measurement performed in a SAC. Good agreements within ±3 dB were obtained over a wide range of frequencies.