Compared to the well-defined C-only process corner models, the situation of RC process corner
models is much more complicated: they are controversial, not well-defined and even design
dependent. In the designs using currently advanced technologies, coupling capacitance between
wires on the same metal layer is dominant (usually more than 70% to 80% of the total
capacitance). The dependency of this coupling capacitance on metal thickness and width is
almost exactly opposite to the metal resistance, resulting in very little change of R*C even when
R and C have individually changed significantly.
This paper will describe how National successfully dealt with the process corner models of RC
extraction in several designs with 0.13ì m technology and how the advanced features
introduced in the latest StarRCXT versions facilitated our methodology implementation. Our
experience showed that rigid uniform parasitic corner models similar to STA approach should
NOT be used. In stead, with the same technology, different designs may need different parasitic
corner models. Furthermore even in the same design, analog parasitic corner models could be
different from its digital counterparts.
Using the StarRCXT ITF file as an example, the paper will also show why and how some
systematic dependences, such as metal width dependency on spacing, metal thickness
dependency on density, etc, should be combined with the process variation corners, in order to
make the RC corner models better reflect real silicon. The paper will share our success
experience as well as our lessons with the Synopsys tool users’ community.
|
Editorial
Upcoming Events
DAC 2012 at San Francisco CA - Jun 3 - 7, 2012
|
|
|
