Skip to Main Content
The use of CFD for thermal analysis in the design of electronic devices has proven to be cost effective and time saving in the ever decreasing product development life cycle, However when the product involves components with several orders of magnitude in length scales, and comprises many fine features which may not be easily simplified because of their random nature, the ability to predict an accurate peak temperature may be called into question. In this paper, a multi-level approach, using compact modeling techniques at the larger scale level, is proposed to deal with the problem of handling fine features within a board or component level simulation, without the needs of a super computer. But increasing detail of the features is represented at the smaller scale levels while maintaining consistency in the boundary conditions between the larger and smaller scale models. This methodology not only identifies the hot spot locations and anticipates the heat distribution pattern, but can also predict peak temperatures within the fine features with more accuracy which are essential for failure analysis of critical features. Using a test vehicle which comprises numerous microvias subjected to different powering modes, it has been shown that the predicted peak temperatures in the critical features can rise by over 12degC as the level of the detail increases. This illustrates the importance of using the multi-level approach when dealing with thermal analysis involving several orders of magnitude in length scales.