|
這應該是APR的論文
% y6 M$ w5 A1 e. [. k5 w9 @
# J7 j0 J+ q* o L, N1 c% } X2 [# f! h
Abstract:+ X( z: q7 m% c& i$ y1 g/ N
Parasitic interconnect corner methods are known to
! \& q) O; \- l, ]. k: k2 ? fbe inaccurate. This paper explains the sources of their errors and$ q& j# k$ ]! u1 x2 A
shows that errors in excess of 22% can occur in the predicted
2 J* |' t# e8 s Y _4 Ncorner delays of a multi-layer stage in the presence of process
. Q) J2 P0 ]4 `variations. It is shown that exhaustive corner search methods are8 c' U% E0 t, {+ k, I1 d
infeasible in practice as they have an exponential complexity in
7 T% n: R0 b! X3 x8 Nterms of required SPICE simulations with respect to the number! t: I* X/ R% M; E* z! h+ X
of layers a stage is routed through. This exponential complexity
8 Y7 L, S; ^8 ]8 Zis reduced to a linear one with a new simulation-based search
; M% R. u1 |( V6 W. S: o; fmethod with the aid of stage delay properties. The ideas behind
4 Y6 m# Y. y; h, s) |the simulation-based methodology are shown to be expandable: j4 w6 m' Q: k8 Y |2 x
to an analytical-based multi-layer performance corner location2 [/ }! I# h3 z4 O( t* |# G! x6 _5 s
methodology. The simulated best/worst case delays based on these. j0 Q4 ~. ~7 |9 @; R. M# }' C/ {
analytical corners produce errors below 4% as compared to the. T- H% W* y0 p1 K
exhaustive search simulation based method.3 w3 C, m3 p8 o5 ~
5 m8 p" V' \" G
[ 本帖最後由 masonchung 於 2008-4-22 12:01 AM 編輯 ] |
本帖子中包含更多資源
您需要 登錄 才可以下載或查看,沒有帳號?申請會員
x
|