Improving the Characteristics of Integrated EMI Filters by Embedded Conductive Layers

Abstract
Discrete electromagnetic interference (EMI) filters have been used for power electronics converters to attenuate switching noise and meet EMI standards for many years. Because of the unavoidable structural parasitic parameters of the discrete filter components, such as equivalent parallel capacitance (EPC) of inductors and equivalent series inductance (ESL) of capacitors, the effective frequency range of the discrete filter is normally limited. Aiming at improving high frequency performance and reducing size and profile, the integrated EMI filter structure has been proposed based on advanced integration and packaging technologies , . Some improvements have been made but further progress is limited by EPCs of the filter inductors, which is restricted by dimension, size and physical structure. In this paper, a new structural winding capacitance cancellation method for inductors is proposed. Other than trying to reduce EPCs, a conductive ground layer is embedded in the planar inductor windings and the structural capacitance between the inductor winding and this embedded layer is utilized to cancel the parasitic winding capacitance. In order to obtain the best cancellation effect, the structural winding capacitance model of the planar spiral winding structure is given and the equivalent circuit is derived. The design methodology of the layout and area of the embedded ground layer is presented. Applying this method, an improved integrated EMI filter is designed and constructed. The experimental results show that the embedded conductive layer can effectively cancel the parasitic winding capacitance, hence ideal inductor characteristics can be obtained. With the help of this embedded conductive layer, the improved EMI filter has much smaller volume and profile and much better characteristics over a wide frequency range, compared to the former integrated EMI filter and the discrete EMI filter.

This publication has 7 references indexed in Scilit: