Experimental determination of thermal conductivity of printed wiring boards

Abstract

Infrared microscopy has been used to measure the thermal conductivity K of printed wiring boards with embedded layers of Cu, for heat flow perpendicular (K/sub /spl perp//) to the plane of the board. The technique provides the thermal conductivities of the individual glass-epoxy layers and the thermal resistance at each interface. It is found that K/sub /spl perp// (glass-epoxy)=0.29 Wm/sup -1/ K/sup -1/, more than three orders of magnitude smaller than that of Cu. The results also show that the bond between Cu and glass-epoxy is sufficiently good that it introduces negligible thermal resistance. A heated-bar technique is used to measure the board-averaged in-plane thermal conductivity (K/sub /spl par//) of many samples of printed wiring board which differ in the amount of circuitry and the number of embedded layers of Cu. Large variations in K/sub /spl par// are found to be uncorrelated with the amount of circuitry or the total number of embedded layers. However, K/sub /spl par// is found to be directly related to the amount of Cu in continuous layers, showing that most of the heat by far is carried by those continuous Cu layers. This leads to a large anisotropy (K/sub /spl par///K/sub /spl perp//) in the board-averaged conductivities, which can be as high as-100 for a board with 2 embedded layers and proportionately higher for more. Such large anisotropy may be an important consideration for reliable thermal management.