|
Moving to nanometer-scale integrated
circuits profoundly affects the entire electronics
industry while creating tremendous
technical challenges. As designs are
getting faster and denser, the issues in
physical design, signal integrity, timing,
and power, become more critical. As a
result, higher performance and more accurate
modeling is required where the extraction
of interconnect models must be
efficient enough to deliver electromagnetic
(EM) accuracy for structures of realworld
complexity.
To address these challenges, researchers
at the University of Manitoba initialized a
research program in which they will develop
efficient modeling methods for the
prediction of signal ropagation in radiofrequency
integrated circuits (RFIC) and
high-speed digital interconnects of submicron
dimensions.
|
 |
Pertaining to efficient
algorithms of computational electromagnetics
(CEM), a new class of methods will
be investigated combining wellconditioned
formulations of integral equations
(IE) with O(NlogN) matriximplicit
computational algorithms. This concurrence
of methods is critical for the construction
of EM solvers with fully controlled
approximation error and capacity
for large-scale applications.
Researchers will explore the parallel implementation
of the algorithms with the
goal of attaining solutions for problems
with up to 20 million unknowns in dense
matrix approximations. This work will
give researchers a unique set of tools for
design automation capable of extracting
interconnect models of unprecedented
sizes such as, for example, the entire RF
module of a radar-on-chip system. Without
the access to computing resources
made available through network connections
afforded by MRnet this type of research
would not be possible.
|
|
