This paper presents a new approach to modelling radiated emissions from electronic systems inside EMC compliance chambers. The proposed approach exploits the fact that manufacturers prefer reusing circuits/layouts in new products with minor geometrical modifications to the existing designs. The available data from previous EMC testing, therefore, may contain certain features/patterns that may be extracted and employed to predict EMC performance of similar products. Motivated by this observation, the proposed method develops accurate models for radiated emissions inside the EMC testing environment. Specifically, radiated emissions from various geometrical configurations of a PCB-cable structure are employed to develop artificial neural network models. These developed models are tested to predict radiated emissions from a given PCB-cable structure configuration. Results show that very accurate estimates of the radiated emissions can be made using the developed models without requiring the time-consuming simulations and expensive prototype testing in the compliance chambers.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below