Fast Prediction of Temperature Evolution in Electronic Devices for Run-Time Thermal Management Applications
Increase of non-uniform power density and high switching frequency has presented new challenges in predicting transient temperature response to fast-changing power inputs in advanced electronic devices. While the computational effort with direct calculation through the finite element model (FEM) is expensive, various methods of model reduction with drastically improved computing speed have been developed for calculation of dynamic thermal responses of the electronic systems. However, those methods’ still-considerable computational time consumption inhibits their practices in real-time temperature prediction and dynamic thermal management (DTM) applications. This work presents a fast algorithm for predicting temperature evolution in electronic devices subjected to multiple heat source excitations. It utilizes the equivalent thermal RC network for model reductions, and adopts recursive infinite impulse response (IIR) digital filters for accelerated computation in discrete time-domain. The algorithm is validated by comparison to existing convolution integral methods, yielding excellent agreement with several orders of magnitude improvement in computation efficiency. Due to its simplicity in implementation, the algorithm is very suitable for run-time evaluation of temperature response for dynamic power management applications.