Computation, modeling and simulation in engineering is based on a long-apparent dichotomy. On the one hand, there are unifying principles of computation, applied math, and physics that pervade most engineering fields. On the other hand, there are domain-specific specialities and advances that impact silo-based disciplines. The ACE lab aims to reconcile this dichotomy by creating innovations in a two-fold manner. Robust, rigorous foundations based on unified principles are developed. These are then fine-tuned, enhanced, and applied in great detail to specific application areas. During this process, new methodologies, interdisciplinary techniques, and paradigm-shifting computational approaches emerge.
The mission of the lab is to be consistent with the emerging long-term evolutionary trend, throughout industry and academia, to return once more to a unified, holistic approach to science and engineering while exploiting the advances made in disciplinary silos. This is an exciting time for such research, and I am always interested in students, collaborators, and sponsors who are interested in advancing algorithms, implementations, and applications along these lines. When possible, the ACE lab also stresses on intellectual property development, licensing, and commercialization, as evidenced in the strong ties with the University of Washington tech transfer office and startup activity - I founded Physware Inc ., a venture-backed electronic design automation and physics simulation startup in 2007.
The strength of the research pursued at the ACE Lab is predicated upon a deep and strong understanding of efficient scalable solutions for classes of equations that arise in electromagnetics. These in turn affect the solution of Laplace, Poisson, Hemlholtz, and Schrodinger equations, and wave-like equations arising in a multitude of applications including computational electromagnetics, photonics, high-speed electronic circuits and packaged systems, biomolecular systems and biochip devices, multi-physics effects in the thermal, fluidic, and elastic arenas, and quantum physics. At the edge of the application space are emerging areas in graph theory and networks, animation, and graphics.
The past and present work and computational resources at the ACE lab have been made possible by several sponsors and funding agencies including the Defense Advanced Projects Research Projects Agency, the National Science Foundation, Semiconductor Research Corporation, the Washington Research Foundation, the Department of Defense SBIR Program, the National Aeronautics and Space Adminstration, Intel Corporation, IBM Corporation, Lawrence Livermore National Labs , and the State of Washington.
More information about my work and background is here . Here is my blog .
Eric Johnson (MS 2000) joined AT&T. Todd West (MS 2001) joined Microsoft. Yong Wang (PhD 2004), co-advised with Prof. Richard Shi, joined Synopsys. Dipanjan Gope (PhD 2005) joined Intel Santa Clara, and then Physware in 2007. Swagato Chakraborty (PhD 2005) joined Applied Wave Research and then Physware in 2007. Chuanyi Yang (PhD 2005) joined Cadence Design Systems. Devan Williams (MS 2007) joined Physware. Dr. Indranil Chowdhury (PhD 2007) will be joining Cadence Design Systems. Dr. Xiren Wang is a postdoctoral research associate, and Ritochit Chakraborty, Ying Li, Mosin Mondal, Venkat Naidu, Joe Peach (co-advised with Prof. Lih Lin), James Pingenot, Arun Sathanur, and Ani Siripuram are graduate students.
Boundary Element Methods are the main class of solution techniques studied and advanced at the ACE Lab. Extensive research in this area by my students have led to complete solver architectures tuned to multiple applications. Specific challenges including low-frequency issues, preconditioning, hierarchical coupling, and detailed applications have been presented.
Fast Multilevel and Multiscale Techniques are the cornerstone of fast integral-equation solvers developed at the ACE lab. Over the past several years, the lab has developed various versions of low-rank solution methods for applications in parasitic extraction, full-wave modeling, and recently in mixed-physics modeling. In addition to low-rank techniques that accelerate iterative methods, we have examined approaches to fast direct solvers. Also, a novel fast multipole method for operating on arbitrary power-law potentials was demonstrated with applications in elastic equations, microfluidic flow, and molecular modeling.
Coupling Electromagnetics and Circuits is critical for hierarchical modeling of complex electronic circuits. My students and I have developed generalized schemes for co-simulation, in both frequency and time domain, of integral equations and circuit equations.
Transient and Temporal Simulation techniques are important for broadband and non-linear systems. We have developed time domain integral equation implementations, including detailed material loss modeling, and have coupled these to SPICE solvers for hierarchical modeling.
Parallelized and Multicore Algorithms are essential in order to scale boundary element technology with advances in muklticore architectures and in the prevalence of low-cost clusters and hybrid architectures. We recently demostrated how accelerated boundary element methods can be mapped onto shared memory architectures in a thread-safe and minimal redundancy manner, and also showed how to do the same for hybrid shared-distributed memory systems using both OpenMP and MPI. These algorithms enable very large scale computational electromagnetics problems to be solved with very high accuracy in reasonable run times, thereby enabling design as well as rapid analysis.
Unified Approaches to Parametric and Statistical Modeling enable accelerated design in the presence of manufacturing tolerances so called Design for Manufacturing (DFM). Rapid response surface generation with minimal calls to expensive field solvers are being developed, with the ability to perform variability, sensitivity, optimization, and statistics on designs. These enable component and sub-system yield prediction of parametrized designs.
Integrated Solver Architectures are necessitated in order to further the specific advances in a manner such that these can be applied to real systems and eventually produce usable software.
Electromagnetic Applications range from scattering from large determinsitic and random structures, to antenna modeling, and design, and EMI/EMC prediction in large platforms.
At the microsystem level,Signal and Power Integrity are of particular relevance.
In emerging communcation and computing technologies,Integrated modeling of RF, Analog, and Mixed-signal Systems in the presence of design and manufacturing variability and the presence of designed and parasitic coupling plays a crucial role.
In lower power applications, passive technologies and in particular RFID Design is now emerging as a strong enabler of new kinds of networks and sensing methodologies,
Moving towards novel architectures and physics, the emergence of the need for point of care in the medical industry has led to Design Tools for Biochips and Labs on Chip as an exciting area.
In other ongoing work, we also look to exploit matrix structure and field behavior in dense and hybrid implict-explicit network modeling and in fast searching and ranking in highly-coupled, user-customized internet environments. Finally, hierarchical petascale design tools based on parametrized quantum and classical solvers for inverse material design, an extremely large-scale and challenging problem, are being investigated.
I will be presenting a four-day short course on computational electromagnetics at Boeing, Seattle on June 17th-20th, 2008. I am also chairing the ICCAD TPC subcommittee on Interconnects and Power Delivery. I am on the TPC of the IEEE EPEP conference where I will present an invited talk at the co-situated Future Directions in IC Packaging workshop. I am on the TPC of the IEC DesignCon conference.
My postdoc and former PhD student Dr. Indranil Chowdhury and I received an inventor's award from NASA for a patent that was filed in the area of high-frequency fast electromagnetic solution.
In addition to technical paper presentations at conferences, my recent talks included invited seminars at Rice University, IBM TJ Watson, National Sun Yat Sen University, a four-day shourt course on computational electromagnetics at Boeing, an invited embedded tutorial on parallel computing methods at IEEE EPEP Atlanta, a contributed tutorial at IEEE DesignCon Santa Clara where I also organized and chaired a panel on chip-package co-simulation, and a technical tutorial presentation at EDSFair Yokohama. I recently presented the UW EE Colloquium seminar.
Dr. Indranil Chowdhury, a postdoc and previous PhD student at the ACE lab, was selected by the NSF in an elite list of 8 students who were asked to present their work in microfludics in a special forum at the ICNMM07 conference in Puebla, and also received an NSF Travel scholarship for the same.
Mosin Mondal, a PhD student in the ACE lab, recently received two honors at conferences: a best paper award at IEEE EPEP Atlanta, and a best paper nomination at IEEE DESIGNCON. He also received a departmental nomination for the IBM graduate fellowship.
Arun Sathanur, a PhD student in the ACE lab, received a departmental nomination for the Intel graduate fellowship.
Joe Peach, a PhD student co-advised with Prof. Lih Lin, received a TGIF grant and a WRF gift for his work on fabrication of a novel DEP-FFF device.
This
Web Page Created with PageBreeze
Free Website
Builder