Delphi style compact modeling for multi-chip package including its bottom board area based on genetic algorithm optimization

Emerging packages offer diverse options for mounting active and passive devices in a single package, such as side by side, embedded or stacked die placements. These high density packages have increased the cooling issues and are demanding more complex board designs to transfer the heat from the chips to the PCB ground planes. The creation of the most compliant behavioral models to a set of DELPHI boundary conditions has been investigated using the Genetic Algorithm fitting technique. The paper describes the comparison of compact and detailed models, dedicated to System in Package component, and the need to simulate in more detail the package and its chips, as well as, the layers, via and μvia of its board mounting area. The thermal behavior predictions, found to be within ±5% of error, show the compliance of Genetic Algorithm for reducing complex multi-chip packages and the feasibility of extending the "Boundary Condition Independence" methodology to the modeling of local board areas. In addition, the use of "local board compact thermal model" linked to each critical component proposes a new way to achieve a more efficient and accurate board cooling optimization.