INVENTIVE
CO
NFI
DEN
TIA
L
DfM
: Wha
t, W
hy, W
hen
& H
ow
Nic
khil
Jaka
tdar
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
2
Ove
rvie
w:
•W
hat i
s D
fM?
•W
hy D
fM?
–Th
e re
aliti
es b
elow
130
nm–
Yie
ld /
Cos
t Ana
lysi
s–
Rul
es a
nd M
odel
s
•W
hen
is it
goi
ng to
be
need
ed?
–E
xam
ples
in C
MP
, Lith
ogra
phy,
and
CA
A
•H
ow is
it g
oing
to b
e so
lved
?–
Nov
el a
ppro
ache
s in
des
ign
impl
emen
tatio
n
•O
bser
vatio
ns a
nd P
redi
ctio
ns
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
3
Wha
t is
Des
ign
for M
anuf
actu
ring
(DfM
)?
•V
ario
us d
efin
ition
s of
DfM
incl
ude
–A
nyth
ing
and
ever
ythi
ng re
latin
g to
a m
ore
mfg
. aw
are
desi
gn o
f a
chip
–R
ET/
OP
C–
Lith
ogra
phy/
CM
P P
roce
ss C
heck
s–
Crit
ical
Are
a A
naly
sis
–S
tatis
tical
Sta
tic T
imin
g A
naly
sis
(SS
TA)
–M
anuf
actu
ring
Aw
are
Rou
ters
–D
esig
n fo
r Mar
ketin
g–
Des
ign
for M
oney
–A
nyth
ing
that
allo
ws
a st
art-u
p to
rais
e a
roun
d of
fund
ing
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
4
Ove
rvie
w:
•W
hat i
s D
fM?
•W
hy D
fM?
–Th
e re
aliti
es b
elow
130
nm–
Yie
ld /
Cos
t Ana
lysi
s–
Rul
es a
nd M
odel
s
•W
hen
is it
goi
ng to
be
need
ed?
–E
xam
ples
in C
MP
, Lith
ogra
phy,
and
CA
A
•H
ow is
it g
oing
to b
e so
lved
?–
Nov
el a
ppro
ache
s in
des
ign
impl
emen
tatio
n
•O
bser
vatio
ns a
nd P
redi
ctio
ns
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
5
Why
Des
ign
mus
t car
e ab
out M
anuf
actu
ring…
•B
elow
130
nm, “
Wha
t yo
u dr
aw is
not
wha
t yo
u ge
t”•
The
furth
er y
ou g
o be
low
130
nm
, the
le
ss a
rule
-bas
ed
appr
oach
is e
ffect
ive
•Th
is le
ads
to d
esig
ns
that
beh
ave
very
di
ffere
ntly
in re
ality
vs
. sim
ulat
ion
little
to
no
pred
icta
bilit
y
Illus
trat
ion
and
phot
omic
rogr
aph
cour
tesy
of T
exas
Inst
rum
ents
.
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
6
Mor
e an
d m
ore
Man
ufac
turin
g Ef
fect
s ca
nnot
be
Mod
eled
with
Rul
es13
0nm
65nm
/45n
m
Via
Failu
re
Part
icle
Def
ects
Lith
oEf
fect
s
CM
PEf
fect
s
Via
Failu
re
Part
icle
Def
ects
Tim
ing,
Pow
erVa
riatio
n
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
7
Impa
ct o
f DfM
on Y
ield
Incr
easi
ng w
ith e
ach
new
pro
cess
gen
erat
ion
0102030405060708090100 0.
35µm
0.25
µm0.
18µm
0.13
µm90
nm
Feat
ure
Dim
ensi
on
Nominal Yields (%)
Proc
ess-
Rela
ted
Des
ign-
Rela
ted
Retic
le-R
elat
ed
Yie
ld σ
: ±5%
Ran
dom
Sys
tem
atic
Sys
tem
atic
Yie
ld σ
: +5%
to -5
0%
Sou
rce:
IBS
(RE
T im
pact
)
(leak
age,
pe
rform
ance
, po
wer
)
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
8
Impa
ct o
f DfM
on D
esig
n C
osts
0.0
20.0
40.0
60.0
80.0
100.
0 0.18
µm0.
13µm
90nm
65nm
Tech
nolog
y No
de
Design Costs ($M)
DFM/
DFY
Costs
Desig
nCo
sts
Obs
erva
tions
:
Des
ign-
spec
ific
yiel
d re
late
d pr
oble
ms
are
grow
ing
Cau
sed
by in
tera
ctio
n of
de
sign
& m
ater
ials
/pro
cess
Incr
easi
ng c
ross
-pro
cess
in
tera
ctio
ns d
iffic
ult t
o m
odel
w
ith ru
les
Mos
t sol
utio
ns li
e be
yond
the
dom
ain
of T
radi
tiona
l DFM
(e
.g. p
ost d
esig
n pr
oces
sing
)
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
9
Impa
ct o
f DfM
on T
ime
to M
arke
t
Effe
ctiv
ity o
f Ite
ratio
nC
ost o
f Ite
ratio
n
Impl
emen
tatio
n
RTL
Sign
off
MFG
$ $$ $$$
$$$$
$$$$
$$$
$$$$
$$$$
$$$$
$
Exp
ecte
d V
olum
e S
hipm
ent
Schedule Delay Act
ual V
olum
e S
hipm
ent
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
10
Trad
ition
al ru
le-b
ased
app
roac
h to
min
imiz
e fa
ilure
s in
des
ign
•A
dd ru
les
Rul
e de
ck s
ize
and
com
plex
ity a
re e
xplo
ding
•M
ake
rule
s m
ore
stric
tH
ighl
y co
nser
vativ
e ru
les
prev
ent b
enef
its o
f ad
vanc
ed p
roce
sses
•A
dd m
argi
nC
ompo
undi
ng e
ffect
of m
argi
n +
mar
gin
+ m
argi
n•
Dra
wba
cks
–S
acrif
ice
perfo
rman
ce–
Incr
ease
are
a +/
or p
ower
–Le
ngth
en d
esig
n cy
cle
–M
iss
unan
ticip
ated
inte
ract
ions
bet
wee
n ef
fect
s
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
11
Des
ign
rule
s ar
e a
trad
eoff
Des
ign
Rul
eS
et Des
ign
Con
stra
ints
Yield
weak
strict
Spee
d vs
. ef
fect
iven
ess
•To
mak
e ru
les
exec
ute
fast
, ke
ep th
em
sim
ple
•To
mak
e si
mpl
e ru
les
effe
ctiv
e,
be c
onse
rvat
ive
and
add
mar
gin
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
12
Cur
rent
app
roac
h: re
com
men
ded
rule
s
Req
uire
dR
ules
Rec
omm
ende
dR
ules Des
ign
Con
stra
ints
Yield
weak
strict
Rec
omm
ende
dru
les
are:
•M
ore
stric
t•
Mor
e co
mpl
ex•
Mor
e ru
n tim
e•
Opt
iona
l
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
13
Mor
e R
ules
–R
ecom
men
ded
Rul
es –
Don
’t So
lve
the
Prob
lem
s
•S
till h
ave
too
man
y ru
les,
too-
stric
t rul
es, t
oo m
uch
mar
gin
•R
ules
don
’t ha
ndle
inte
ract
ion
of e
ffect
s –
mul
tivar
iate
cor
rela
tions
–Li
tho
step
per s
hallo
w d
epth
of f
ocus
↔un
even
chi
p su
rface
from
CM
P–
OP
C o
n ad
jace
nt D
RC
-cle
an s
hape
s pr
oduc
es a
rtifa
cts
(ext
ra s
hape
s)–
Mul
tiple
DR
C-c
lean
laye
rs s
tack
up
toal
low
sof
t spo
t for
CM
P e
rosi
on•
Rul
es a
lway
s co
mpr
omis
e be
twee
nTy
pe I
and
Type
II e
rrors
–Ty
pe I:
mis
sed
a tru
e pr
oble
m–
Type
II: h
ighl
ight
a fa
lse
alar
m•
Rul
es d
on’t
enab
le in
telli
gent
risk
/ben
efit
trade
off
Bot
tom
-Lin
e: R
equi
red
Rul
es v
s. R
ecom
men
ded
Rul
es D
rive
Des
igne
rs C
razy
!!
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
14
Ove
rvie
w:
•W
hat i
s D
fM?
•W
hy D
fM?
–Th
e re
aliti
es b
elow
130
nm–
Yie
ld /
Cos
t Ana
lysi
s–
Rul
es a
nd M
odel
s
•W
hen
is it
goi
ng to
be
need
ed?
•H
ow is
it g
oing
to b
e so
lved
?–
Nov
el a
ppro
ache
s in
des
ign
impl
emen
tatio
n
•O
bser
vatio
ns a
nd P
redi
ctio
ns
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
15
Para
met
ric Y
ield
Ver
y Se
nsiti
ve to
Lith
o Pr
oces
s Va
riatio
ns
•Fo
r a 2
-inpu
t NA
ND
Gat
e:a
chan
ge in
Pol
y G
ate
Leng
th o
f 10
%
–20
% in
del
ay v
aria
tions
(65
NM
)–
100%
in le
akag
e (9
0 N
M).
•W
idth
and
leng
th v
aria
tion
can
be p
redi
cted
by
usin
g M
odel
B
ased
Lith
ogra
phy
Ana
lysi
s:–
At t
he L
ibra
ry D
esig
n le
vel
–A
t the
chi
p le
vel
0.16
0.17
0.18
0.19
0.20
30405060708090
Leakage Current (pA)
Draw
n Ga
te L
engt
h (u
m)
±10%
L
±100
%
I sub
L cd
65 n
m
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
16
Inte
rcon
nect
Var
iatio
ns –
The
Z ax
is
Nar
row
er L
ines
Mea
n La
rger
Impa
ct o
n El
ectr
ical
Par
amet
ers
130n
m >
90n
m >
65n
mC
u lo
ss d
ue to
di
shin
g (o
r ero
sion
)
Dec
reas
ing
Cro
ss-S
ectio
nal A
rea
of L
ine
Add
ition
al M
etal
Lay
ers
Com
poun
d th
e To
pogr
aphi
cal
Impa
ct
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
17
Para
met
ric Y
ield
Ver
y Se
nsiti
ve to
CM
P Pr
oces
s Va
riatio
ns
0.95
5
0.97
7
1.00
0
1.02
3
1.04
5
1.06
8
1.09
1
0.95
5
0.97
7
1.00
0
1.02
3
1.04
5
1.06
8
1.09
1
0.95
50.
977
1.00
01.
022
1.04
51.
068
1.09
10.
955
0.97
71.
000
1.02
21.
045
1.06
81.
091
0.95
50.
977
1.00
01.
022
1.04
51.
068
1.09
1
0.0%
5.0%
10.0
%
15.0
%
20.0
%
25.0
%
30.0
%
35.0
%
40.0
%
45.0
%
01
23
45
67
89
1011
12
Cap
Delta
(%)
Nets (%)
5ps
bette
rM
inor
Path
4
OK
Min
orPa
th3
15ps
wor
seO
KPa
th2
8ps
wor
seO
KPa
th1
Mod
elR
ule
Mod
elR
ule
Hol
d Se
tup
RC
Var
iatio
n
Thic
knes
s V
aria
tion
Tim
ing
Var
iatio
nJo
int p
aper
“Inc
orpo
ratio
n of
CM
P M
odel
ing
in R
C E
xtra
ctio
n an
d Ti
min
g Fl
ow”
byH
. Lia
o, L
. Son
g, N
. Jak
atda
r, R
. Rad
ojci
c) s
ubm
itted
to D
AC
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
18
Man
ufac
turin
g ef
fect
s in
tera
ct (2
)C
MP
prod
uces
big
var
iatio
ns in
sur
face
hei
ght
Com
mon
pro
cess
win
dow
Lith
o ef
fect
+ C
MP
effe
ct=
dist
orte
d / m
issi
ng fe
atur
eyi
eld
failu
reD
oF: +
/-10
0nm
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
19
Impa
ct o
f Sys
tem
atic
Inte
rcon
nect
Var
iatio
ns @
65n
m
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
20
Ove
rvie
w:
•W
hat i
s D
fM?
•W
hy D
fM?
–Th
e re
aliti
es b
elow
130
nm–
Yie
ld /
Cos
t Ana
lysi
s–
Rul
es a
nd M
odel
s
•W
hen
is it
goi
ng to
be
need
ed?
•H
ow is
it g
oing
to b
e so
lved
?–
Nov
el a
ppro
ache
s in
des
ign
impl
emen
tatio
n
•O
bser
vatio
ns a
nd P
redi
ctio
ns
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
21
Phys
ical
Effe
cts
Con
nect
Cor
e to
Adj
acen
cyDesign Methods
Des
ign
CO
RE
Mas
k / M
frg
“Pos
t Des
ign”
AD
JAC
ENC
Y
“Gol
den”
GD
S
Des
ign
Mas
k / M
frg
“Gol
den”
GD
S“G
olde
n”G
DS
Layo
ut O
ptim
izat
ion
Para
sitic
Ext
ract
ion
LVS/
DR
C
Sign
-Off
PVB
atch
RET
Tre
atm
ent
Verif
icat
ion
(OPC
, CM
P …
)
P+R
Layo
ut
Litho
CMP
Etch
Yiel
d R
amp
and
FA
Dig
ital S
oCC
usto
m
Indu
stry
Stru
ctur
e
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
22
Des
ign
Mas
k / M
frg
“Gol
den”
GD
S“G
olde
GD
S
Layo
ut O
ptim
izat
ion
Para
sitic
Ext
ract
ion
LVS/
DR
C
Sign
-Off
PVB
atch
RET
Tre
atm
ent
Verif
icat
ion
(OPC
, CM
P …
)
Yiel
d R
amp
and
FA
Dig
ital S
oCP+
RC
usto
mLa
yout
Litho
CMP
Etch
Electrical Analysis
Physical Analysis
RLC
Mod
el B
ased
Des
ign
& A
bstr
actio
n Li
nks
Impl
emen
tatio
n to
Man
ufac
turin
g
n”
Indu
stry
Stru
ctur
e
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
23
Con
text
-bas
ed, A
dapt
ive
Mod
el R
esol
utio
n
•In
the
Des
ign
–In
crea
sed
mod
el
reso
lutio
n fo
r opt
imiz
ing
criti
cal a
nd s
ensi
tive
path
s
•In
the
Flow
–A
dapt
ive
mod
el re
solu
tion
and
spee
d-ac
cura
cy tr
ade-
off t
o m
atch
abs
tract
ion
leve
lR
TL
Synt
hesi
s
Pro
toty
ping
Phy
sica
l Sy
nthe
sis
Rou
ting
Opt
imiz
atio
n
Sig
n-of
f
Model Resolution
Net
s/Pa
ths
Reg
ions
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
24
Cus
tom
Des
ign
Met
hods
Evo
lutio
n
Imag
e In
tens
ity M
ap
Com
mon
pro
cess
win
dow
Det
erm
ine
trans
isto
r mat
chin
g to
pro
cess
var
iatio
n of
dos
e an
d fo
cus
by e
xplo
ring
com
mon
pro
cess
win
dow
on
sele
cted
are
as /
cutli
nes
Silic
on im
age
of m
ultip
le p
ost t
reat
men
tlay
outs
ana
lyze
d to
de
term
ine
sens
itivi
ty o
f orig
inal
desi
gn la
yout
with
imag
e in
tens
ity m
ap
25
Dur
ing
Impl
emen
tatio
nIn
terc
onne
ct O
ptim
izat
ion/
Lith
o R
epai
r
•Low
per
turb
atio
n w
ith
cont
rolla
ble
elec
trica
l ch
arac
teris
tics
•Fas
t and
pre
cise
with
hig
h fix
rate
s
•Inc
rem
enta
l tim
ing
anal
ysis
Cas
e 1:
Slid
e vi
a do
wn
or m
ove
wire
-20
nm
Chi
p O
ptim
izer
Exa
mpl
es
Lith
o er
rors
Lith
o m
odel
Fixi
ng g
uide
lines
Cas
e 4:
B
efor
eA
fter (
wire
/via
mov
e)
Cas
e 3:
B
efor
eA
fter (
bend
/topo
opt
imiz
atio
n)
Cas
e 2:
Wid
en jo
gs
Cad
ence
Chi
p O
ptim
izer
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
26
Smar
t rou
ting
prev
ents
yie
ld p
robl
ems
•N
ew la
yout
rule
s an
d gu
idel
ines
•R
edun
dant
via
s an
d re
duce
d vi
a co
unt
•W
ire s
prea
ding
Via
optim
izat
ion
Wire
spr
eadi
ng
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
27
Dur
ing
Impl
emen
tatio
nLi
tho
Aw
are
Rou
ting
Con
vent
iona
l P&
R
050100
150
200
250
12
34
56
Diff
eren
t Rou
ting
# of Necking Defects
Lith
o-A
war
e P
&R
SoC
Enco
unte
r
Febr
uary
16,
200
7C
aden
ce C
onfid
entia
l: C
aden
ce In
tern
al U
se O
nly
28
DfM
Impl
emen
tatio
n: K
ey T
ake-
Aw
ays
•M
odel
s ar
e re
quire
d to
redu
ce g
uard
ban
d by
impr
oved
m
odel
ing
of s
yste
mat
ic e
ffect
s•
A p
latfo
rm fo
r con
curr
ent a
naly
sis
and
optim
izat
ion
of
effe
cts
is re
quire
d fo
r con
verg
ence
•In
crem
enta
l ana
lysi
s is
requ
ired
for o
ptim
izat
ion
•M
argi
n dr
iven
mod
el a
ccur
acy
is n
eede
d to
focu
s co
mpu
tatio
nal r
esou
rces