Thermal Model Simplification of Mobile Device with Adaptive Metamodel of Optimal Prognosis (AMOP)

Abstract

The effective performance of system on chip (SOC) is tightly coupled with transient thermal behavior. To consider multiple design parameters in the limited development time budget, a simple and accurate thermal model is required. Conventional workflow is inefficient to develop equivalent simple set models and tune the thermal properties with trial and error. We introduce the workflow to develop a simplified thermal model using computational fluid dynamics (CFD) solver driven by process integration and design optimization (PIDO) platform. The process contains design of experiment (DOE), sensitive analysis, advanced metamodel of optimal prognosis (AMOP), evolution algorithm (EA), and optimization. Smartphone model with application processor (AP) was used for a proof of concept. By using the AMOP + EA approach, a simple thermal model was developed successfully within acceptable error. To verify the benefit of this method, the best simple model from optimization tool and full set model were compared for two different types of AP packages, package on package (POP) and single chip package (SCP).