Elmore Delay, Logical Effort

Elmore Delay, Logical Effort
Modern Interconnect
© Rabaey, ch4Wire.ppt, slide 22

Example: Intel 0.25 micron Process
5 metal layers
Ti/Al - Cu/Ti/TiN
Polysilicon dielectric

Modern Interconnect
• 90nm process
© Chris Kim (image from Intel?)

The Lumped RC-Model, The Elmore Delay
(result *0.69)

Example: The Elmore Delay
Shared Paths:
R44 = R1+R3+R4
Rii = R1+R3+Ri
Ri4 = R1+R3
Ri2 = R1
Ti = C1R1 + C2R1 + C3(R1+R3) + C4(R1+R3) + Ci(R1+R3+Ri)

The Elmore Delay RC Chain
© Rabaey, ch4Wire.ppt
The Distributed RC-line
Diffusion
Equation
Deriving the Diffusion Eq
Step-response of RC wire as
a function of time and space
2.5
x= L/10
2
x = L/4
voltage (V)
1.5
x = L/2
1
x= L
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
time (nsec)
RC-Models
Driving an RC-line
Rs (r w,cw,L)
Vout
V
in
Designing Fast CMOS Gates
Slides from chapter6.ppt of

Rabaey’s page
Fan-In Considerations
A B C D
A CL Distributed RC model
B C3 (Elmore delay)
C C2 tpHL = 0.69 Reqn(C1+2C2+3C3+4CL)
D C1
Propagation delay deteriorates
rapidly as a function of fan-in –
quadratically in the worst case.
© Digital
EE141 Integrated Circuits2nd
Combinational Circuits
tp as a Function of Fan-In
1250
quadratic
1000
Gates with a
750
fan-in
tp (psec)
tpH tp greater than

500
L
4 should be
250 tpL avoided.
H linear
0
2 4 6 8 10 12 14 16
fan-in
© Digital
tp as a Function of Fan-In and Fan-Out
 Fan-in: quadratic due to increasing

resistance and capacitance
 Fan-out: each additional fan-out gate
adds two gate capacitances to CL
tp = a1FI + a2FI2 + a3FO
© Digital
Fast Complex Gates:
Design Technique 1
 Transistor sizing
 as long as fan-out capacitance dominates
 Progressive sizing
Distributed RC line
InN MN CL
M1 > M2 > M3 > … > MN
(the fet closest to the
In3 M3 C3 output is the smallest)
In2 M2 C2 Can reduce delay by more than

In1 20%; decreasing gains as
M1 C1
technology shrinks
© Digital
Fast Complex Gates:
Design Technique 2
 Transistor ordering
critical path critical path
charged 01
In3 1 M3 CL In1 M3 CLcharged
In2 1 M2 In2 1 M2 C2 discharged

C2 charged
In1 In3 1 M1 C1 discharged
M1 C1 charged
01
delay determined by time to delay determined by time to

discharge CL, C1 and C2 discharge CL
© Digital
Fast Complex Gates:
Design Technique 3
 Alternative logic structures
F = ABCDEFGH
© Digital
Fast Complex Gates:
Design Technique 4
 Isolating fan-in from fan-out using buffer
insertion
CL CL
© Digital
Logical Effort
Slides from chapter6.ppt of

Rabaey’s page
Transistor Sizing
Cg=
Cint= Cg= Cint=
Cint= Cg=
D=1+f D=2+4/3 f D=2+5/3 f
© Digital
Normalized Space
Parasitic Term P
NOTE: p is a gate parameter  function(W)

Logical Effort Term g
NOTE: g is a gate parameter  function(W)

Transistor Sizing a Complex
CMOS Gate
B 8 6
A 4 3
C 8 6
D 4 6
OUT = D + A • (B + C)
A 2
D 1
B 2C 2
© Digital
Logical Effort
From Sutherland, Sproull
© Digital
Logical Effort of Gates
2
=
p
3;
4/
5 1
=
p=
g
Normalized Delay
D:
4 AN 1;
g =
:
tN
e r
ert
pu
3 v
in
In Effort
2-
Delay
2
1
Intrinsic
Delay
1 2 3 4 5
Fanout f
© Digital
tp as a Function of Fan-Out
All gates
tpNOR2 tpNAND2 have the
same drive
tpINV current.
tp (psec)
Slope is a
function of
“driving
strength”
2 4 6 8 10 12 14 16
eff. fan-out
© Digital
Buffer Example
In Out
1 2 N CL
N
Delay    pi  g i  f i  (in units of tinv)
i 1
For given N: Ci+1/Ci = Ci/Ci-1

To find N: Ci+1/Ci ~ 4
How to generalize this to any logic path?
© Digital
Delay in a Logic Gate
Gate delay:
d=h+p
effort delay intrinsic delay
Effort delay:
h=gf
logical effective fanout =

effort Cout/Cin
Logical effort is a function of topology, independent of sizing
Effective fanout (electrical effort) is a function of load/gate size
© Digital
Add Branching Effort
Branching effort:
Con path  Coff  path

b
Con path
© Digital
Multistage Networks
N
Delay    pi  g i  f i 
i 1
Stage effort: hi = gifi

Path electrical effort: F = Cout/Cin
Path logical effort: G = g1g2…gN
Branching effort: B = b1b2…bN
Path effort: H = GFB
Path delay D = Sdi = Spi + Shi
© Digital
Optimum Effort per Stage
When each stage bears the same effort:

hN  H
hN H
Stage efforts: g1f1 = g2f2 = … = gNfN
Effective fanout of each stage: f i  h g i
Minimum path delay
Dˆ   gi f i  pi   NH 1/ N  P
© Digital
Optimal Number of Stages
For a given load,
and given input capacitance of the first gate
Find optimal number of stages and optimal sizing
D  NH 1/ N  Npinv
D
N
 
  H 1/ N ln H 1/ N  H 1/ N  pinv  0
1/ Nˆ
Substitute ‘best stage effort’ hH
© Digital
Example: Optimize Path
1 b c
a
5
g=1 g = 5/3 g = 5/3 g=1
f=a f = b/a f = c/b f = 5/c
Effective fanout, F = 5
G=
H=
h=
a=
b=
c=
© Digital
Example: Optimize Path
1 b c
a
5
g=1 g = 5/3 g = 5/3 g=1
f=a f = b/a f = c/b f = 5/c
Effective fanout, F = 5
G = 25/9
H = 125/9 = 13.9
h = 1.93
a = 1.93
b = ha/g2 = 2.23
c = hb/g3 = 5g4/f = 2.59
© Digital
Example – 8-input AND
© Digital
Method of Logical Effort
 Compute the path effort: F = GBH

 Find the best number of stages N ~ log4F
 Compute the stage effort f = F1/N
 Sketch the path with this number of stages
 Work either from either end, find sizes:
Cin = Cout*g/f
Reference: Sutherland, Sproull, Harris, “Logical Effort, Morgan-Kaufmann 1999.
© Digital
Summary
Sutherland,
Sproull
Harris
© Digital

Elmore Delay, Logical Effort

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Elmore Delay, Logical Effort

Uploaded by

Copyright:

Available Formats

Elmore Delay, Logical Effort

© Rabaey, ch4Wire.ppt, slide 22

© Rabaey, ch4Wire.ppt, slide 23

© Chris Kim (image from Intel?)

© Rabaey, ch4Wire.ppt, slide 27

Ti = C1R1 + C2R1 + C3(R1+R3) + C4(R1+R3) + Ci(R1+R3+Ri)

Slides from chapter6.ppt of

tpH tp greater than

 Fan-in: quadratic due to increasing

tp = a1FI + a2FI2 + a3FO

In2 M2 C2 Can reduce delay by more than

In2 1 M2 In2 1 M2 C2 discharged

delay determined by time to delay determined by time to

Slides from chapter6.ppt of

D=1+f D=2+4/3 f D=2+5/3 f

NOTE: p is a gate parameter  function(W)

NOTE: g is a gate parameter  function(W)

From Sutherland, Sproull

For given N: Ci+1/Ci = Ci/Ci-1

logical effective fanout =

Con path  Coff  path

Stage effort: hi = gifi

Path delay D = Sdi = Spi + Shi

When each stage bears the same effort:

Effective fanout of each stage: f i  h g i

Minimum path delay

 Compute the path effort: F = GBH

Reference: Sutherland, Sproull, Harris, “Logical Effort, Morgan-Kaufmann 1999.

You might also like