|
|
Magic Mailing List |
|
From: bindu madhavan (madhavan AT usc DOT edu) Date: Tue Jul 30 2002 - 21:46:10 EDT
Hi,
I am seeing the following problem with extraction of
mosfet perimeter values extracted in magic-7.1.x
both from the magic-7.1.2 distribution and the CVS
repository as of July 10, 2002.
The problem is independent of tech file,
and observed on the following platforms: HP
workstation running HPUX 11.0, magic ported to cygwin
on i686 platform)
Problem statement (ext_problem.txt in attched gz
file):
magic-7.1.x extraction has incorrect PD and PS values
for NMOS, PMOS
transistors, when there is more than one
rectangle involved in
the description of the shape of diffusion --
for example: diffusion
contact, or right-angle turn of diffusion.
Problem does not exhibhit itself in magic-65.
Problem example (to highlight problem, see nmos.mag
file in attached gz file):
NMOS transistor 6 lambda wide, with diffusion overlap
of gate of
5 lambda. Source has N-Diff contact 6-lambda wide and
4-lambda
tall. Drain has no N-Diff contact.
Correct answer should be:
.option scale=......
m1 drain gate source bulk nfet w=6 l=2 ad=30 pd=22
as=30 ps=22
instead, I get:
m1 drain gate source bulk nfet w=6 l=2 ad=30 pd=22
as=30 ps=34
The difference, it seems, is that magic-7.1.x is
considering
the source perimeter to be comprised of the
following:
6 lambda x 1 lambda piece of n-diff (compute peri = 2
( 6 + 1) = 14)
6 lambda x 4 lambda piece of ndc (compute peri = 2 (6
+ 4) = 20)
to give a total peri = 34 lambda.
Another example would be the second transistor in
nmos.mag, with
terminals, drain2, gate2 and source2.
Any help on this matter is greatly appreciated.
Thanks very much in advance
bindu madhavan
madhavan AT usc DOT edu
madbindu AT yahoo DOT com
|
|
|
|