High-precision compact-thermal models

Abstract

Classical approaches proposed for the construction of compact-thermal models suffer from a common disadvantage. They provide no means for error control. Although it is evident that the more nodes we define on the surface, the higher will be the precision, there is no guidance as to the required number of nodes to be defined on the surface, as well as the best position and size to obtain a given precision. Intuitive approaches usually used would fail to deal with complex objects such as those having multiple independent heat sources [multichip modules (MCM)]. In this work, based on an original and generalized formulation of the Green's function, a systematic approach is proposed to build the compact model that gives a better control of these errors, allowing thus the creation of compact models at any desired order of precision. It also gives access to new data concerning tangential gradients on the surface.