Abstract
As the number of chips within electronic packaging increases while their feature sizes shrink, the thermal performance of multichip modules (MCMs) becomes increasingly critical. To upgrade the thermal performance of MCMs without changing package structure and materials, optimal chip placement is an important research aspect. Therefore, this work aims at enhancing the thermal performance of MCMs under natural convection by optimizing the chip placement. To attain the goal, novel chip placement optimization schemes are proposed. The chip placement optimization schemes proposed include the direct optimization scheme, the indirect optimization scheme, the hybrid optimization scheme and the modified hybrid optimization scheme. The direct optimization scheme directly calculates the total chip junction temperature of a MCM under various chip placements by ANSYS® finite element thermal analysis and obtains the optimal chip placement through genetic algorithm evolution. The indirect optimization scheme combines the ANSYS® finite element thermal analysis and a response surface based modified superposition method to effectively construct the response surfaces of chip junction temperature. With the constructed chip junction temperature response surfaces and sequential quadratic programming, an optimal chip placement and total chip junction temperature can be achieved. The hybrid optimization scheme combines the merits of the direct optimization scheme and the indirect optimization scheme. It integrates the ANSYS® finite element thermal analysis, a global search heuristic algorithm, genetic algorithm, and a response surface method to construct the chip junction temperature equations. The chip junction temperature equations constructed are adopted in sequential quadratic programming to find the optimal chip placement. The modified hybrid optimization scheme is the modification of the hybrid optimization scheme. The concept of design space reduction is employed. The procedures of adding new chip placement designs, constructing new chip junction temperature equations and finding minimum total chip junction temperature are repeated to upgrade the efficiency of the hybrid optimization scheme. To demonstrate the effectiveness of the proposed optimizaiton schemes, several thermal dissipation analysis problems associated with two types of MCM are performed. The results obtained show that the direct optimization scheme can only find better solution after massive computations. The response surface based modified superposition method proposed in the indirect optimization scheme can construct chip junction temperature equations that take the influence of chip power and placement into account. It has the same accuracy as the conventional response surface method with less computation. The hybrid optimization scheme largely reduced the computation time toward optimization and its computational efficiency is much higher than the conventional response surface method, the direct optimization scheme and the indirect optimization scheme. However, the constructed response surface is still insufficient to provide accurate total chip junction temperature under the optimal chip placement found. Although the modified hybrid optimization scheme needs a few more computations than hybrid optimization scheme, but its computational efficiency is much higher than the conventional response surface method, the direct optimization scheme and the indirect optimization scheme. It can calculate the optimal chip placement and total chip junction temperature effectively and accurately. Through an appropriate adjustment of the tolerance taken, the modified hybrid optimization scheme is especially useful for larger-scale MCM thermal design problems. With the consideration of electrical requirement, this work can be further devoted to deal with the optimization of chip placement of MCMs considering both thermal and electrical problems. Moreover, by incorporating the current result with thermal stress analysis, the reliability of MCMs can also be analyzed and used as a guideline for the thermal management of MCMs.