CAREER: Full Wave Analysis of Electronic Packaging on High Performance Computers

9624628格德尼近年来，半导体设备的速度和密度大幅提高，产生了具有数以百万计晶体管的单芯片设备、时钟速度接近500兆赫兹的数字设备以及工作频率高达250兆赫兹的微波设备。为了支持这种高速器件，电子封装，即支持半导体器件的电气互连网络，必须小型化，并优化其设计以保持信号的完整性。因此，发展严格的方法来准确地分析和设计电子封装的电气互连网络是至关重要的。这项研究工作将开发全波分析工具，能够分析用于数字电路和微波电路应用的实用大规模电子封装。为此，主要目标是：1)提供密集电子封装的全波电磁场分析能力。2)利用高阶基函数和小波展开，推导出麦克斯韦方程的显式和隐式时间域解。3)利用时间域有限差分方法和有限元时间域方法，建立了麦克斯韦方程的显式/隐式混合解。4)将完全匹配层(PML)边界条件推广到基于非结构网格的显式和隐式时间域格式。5)开发了适用于异构型分布式多处理器计算环境的隐式和显式格式以及显式/隐式混合方法的创新并行算法。6)建立工业分支机构，指导全波仿真工具的实际应用，并为这些工具在设计环境中的应用提供试验台。教学工作最初将集中在对现有电磁学课程的修改上。这项工作的目标是通过使用交互式软件工具箱和开发可供本科生和研究生用于教学目的的案例研究库，向学生提供应用知识。这将使学生能够通过可视化真实世界应用中的物理现象来可视化对电磁场和波相互作用的理解，从而使学生对电磁场和波相互作用有更好的物理理解。此外，通过案例研究，学生将获得更好的电磁学应用知识。最后，多处理器计算机正在达到这样一个点，它们对工业来说变得经济，并将越来越多地用于工程设计。然而，只有少数几名工程师和计算机科学家接受了在这些平台上开发软件的培训。其中一个目标将是提供一门关于并行计算基础的课程，以及使用一些平台和范例进行并行算法开发的实践经验。这门课程将是跨学科的，将为一些大学的工程学学生以及肯塔基州的工业工程师提供关于并行算法开发和并行范例的基础和实践理解。***

9624628 Gedney In recent years, the speeds and densities of semiconductor devices have increased dramatically, resulting in single chip devices with millions of transistors, digital devices operating with clock speeds approaching 500 MHz, and microwave devices with operating frequencies upwards to 250. In order to support such high speed devices, the electronic packaging, namely, the electrical interconnect networks supporting the semiconductor devices, must be miniaturized and their design optimized to maintain signal integrity. Therefore, it is vital to develop rigorous methods to accurately analyze and design electrical interconnect networks of electronic packages. This research effort will develop full wave analysis tools that this is capable of analyzing practical large scale electronic packages for digital circuit and microwave circuit applications. To this end, the principal objectives are: 1) To provide the capability of full-wave electromagnetic field analyses of dense electronic packages. 2) Develop explicit and implicit time-domain solutions of Maxwell's equations specifically using higher-order basis functions, and also utilizing Wavelet expansions. 3) Develop a hybrid explicit/implicit solution of Maxwell's equations using the finite-difference time domain and finite-element time-domain methods. 4) Extend the perfectly matched layer (PML) boundary condition to explicit and implicit time-domain schemes based on unstructured meshes. 5) Develop innovative parallel algorithms for implicit and explicit schemes for heterogeneous distributed multiprocessor computing environments, and for hybrid explicit/implicit methods. 6) Establish industrial affiliates to mentor the practical applications of the full-wave simulation tools and to provide test beds for these tools to be applied in a design environment. The teaching effort will initially focus on the modification existing courses in electromagnetics. The objective of this effort is to provide an applied knowledge to students through the use of interactive software toolboxes and the development of a case study library that can be presented for instructional purposes at the undergraduate and graduate level. This will enable students to have a better physical understanding of electromagnetic field and wave interactions by being able to visualize understanding of electromagnetic field and wave interactions by being able to visualize the physical phenomena in real world applications. Furthermore, through case studies, students will be provided with a better applied knowledge of electromagnetics. Finally, multiprocessor computers are reaching the point where they are becoming economical for industry and will be used more and more in engineering design. However, only a select few engineers and computer scientists are trained on development of software on these platforms. On of the goals will be to provide a course in the fundamentals on parallel computing, with hands on experience of parallel algorithm development using a number of platforms and paradigms. This course will be cross-disciplinary and will provide a foundation and practical understanding of parallel algorithm development and parallel paradigms to engineering students at a number of Universities as well engineers in Industry in the Commonwealth of Kentucky. ***