0% found this document useful (0 votes)
128 views16 pages

4740 Lecture06 Inverter Sizing

This document summarizes key concepts from a lecture on inverter sizing for digital VLSI design. It discusses propagation delay, rise/fall times, contamination delay, using an RC circuit model to calculate delays. It explains how increasing the width of both NMOS and PMOS transistors (the sizing factor S) can reduce propagation delay but increases area. For a chain of inverters, there is an optimal sizing where each inverter is sized up by the same factor compared to the previous gate, to minimize total delay through the chain.

Uploaded by

krrish06230
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
128 views16 pages

4740 Lecture06 Inverter Sizing

This document summarizes key concepts from a lecture on inverter sizing for digital VLSI design. It discusses propagation delay, rise/fall times, contamination delay, using an RC circuit model to calculate delays. It explains how increasing the width of both NMOS and PMOS transistors (the sizing factor S) can reduce propagation delay but increases area. For a chain of inverters, there is an optimal sizing where each inverter is sized up by the same factor compared to the previous gate, to minimize total delay through the chain.

Uploaded by

krrish06230
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

6/8/2018

ECE4740:
Digital VLSI Design
Lecture 6: Inverter sizing

193

Very important concepts

Brief timing recap

194

1
6/8/2018

Rise/fall times & propagation delay


propagation delay
Vin
tp = (tpHL+tpLH)/2
input 50%
waveform

t
tpHL tpLH
Vout
90% 90%

output
waveform 50%

10%
t
tf tr

195

Contamination delay
Vin contamination delay
tcd = 0.5(tcdHL+tcdLH)
input 50%
waveform

t
tcdHL tcdLH
Vout

output tcdHL and tcdLH


waveform 50% can differ

196

2
6/8/2018

First-order analysis: extract Req


• Approximations required as CL(v) and i(v)
are non-linear and voltage v dependent

197

Use RC circuit to model tpHL


• Decay time from 100% to 50% is

• Same can be obtained for low-to-high time

• Propagation delay:

198

3
6/8/2018

Increase both Wn and Wp!

Inverter sizing

199

Split load capacitance CL

intrinsic output
extrinsic
capacitances from
capacitances:
self-loading etc.
fan-out and wiring
(diffusion, Miller)

• Just rewrite as:

“unloaded” propagation delay effect on propagation delay caused


by extrinsic capacitances
assume that NMOS and PMOS are sized
200
such that rise and fall times are equivalent

4
6/8/2018

Impact of sizing factor S

• Intrinsic capacitances:
• Gate resistance: capacitance of a
unit-sized device
resistance of a
unit-sized device

• Leads to:

201

An equivalent view
VDD VDD
Req=Rref/S
high
Vin Vout Vout

Cext Cint= Cref*S Cext


low

VSS VSS

• Propagation delay:
• Simplifying:

202

5
6/8/2018

What exactly is “S”?


• The sizing factor is how much larger you
make the widths of NMOS and PMOS
– You usually keep the PMOS:NMOS ratio fixed
– Gate lengths L usually kept at the minimum

VDD

Vout
Wp S=2 corresponds to
Vin
2Wp and 2Wn
Wn
VSS
203

Impact of sizing S (cont’d)

intrinsic delay
of inverter: tp0

• Intrinsic delay of gate independent of S


– no load means no effect on propagation delay
• Making S large, eliminates effect of load
– at the cost of inverter area!
– and it will affect the load of the preceding logic! 204

6
6/8/2018

Example: Sizing impact* on tp


3.8
x 10-11
3.6
3.4
3.2
3
2.8 self loading
2.6 (intrinsic capacitance
dominates)  not
2.4 much gain anymore
2.2
2
1 3 5 7 9 11 13 15
S
large improvement
already achieved for S=5 Image taken from:
http://bwrcs.eecs.berkeley.edu/Classes/icdesign/ee141_f01/Notes/chapter5.pdf

*for a fixed load! 205

A more relevant case

Sizing a chain of inverters

206

7
6/8/2018

Sizing a chain of inverters

• For isolated inverter: Increasing S reduces


delay but also increases input capacitance
– not very useful in practice!

• More relevant case: chain of inverters:

207

It’s a trade-off
• Sizing up an inverter reduces delay, but will
also increase its input capacitance!
• Intrinsic capacitance Cint proportional to
gate capacitance:

• is technology dependent (and about 1)


208

8
6/8/2018

The effective fan-out f


• Delay of an inverter: a function of the ratio
between external load and its input cap!

• Effective fan out:

• The goal is to minimize the delay through


the entire inverter chain
209

minimum
Chain of N inverters
sized final load

in out
1 2 N
Cg,1 no wire CL
loads

• Delay of jth inverter:


• Total delay is:
• Assume:

• Optimality conditions:
210
Image adapted from: Digital Integrated Circuits (2nd Edition) by Rabaey, Chandrakasan, Nikolic

9
6/8/2018

Optimum size of inverter

• Optimum size of each inverter is


geometric mean of neighbor’s sizes!

211

Implications

• Each inverter should be sized up by same


factor fopt compared to preceding gate

WHY?

212

10
6/8/2018

Implications (cont’d)

• With

the minimum (total) delay is

• Simply a result of

213

Example: simple chain


in
out

1 f f2 C L= 8 C 1
C1

• F=CL/C1 has to be evenly distributed


across N=3 stages:

214

11
6/8/2018

Optimal number of stages Nopt

• There is a trade-off
– too many stages: intrinsic delay dominates
– too few stages: effective fan-out dominates

• Has no closed-form solution 


• Solution is about 3.6 for
215

Assume no self loading

• Implies
• Solve

• Leads to so-called exponential horn

216

12
6/8/2018

Optimum effective fan-out fopt


5 7

6
4.5
5
4
4

3.5 3

2
3
1
2.5 0
0 0.5 1 1.5 2 2.5 3
1 1.5 2 2.5 3 3.5 4 4.5 5
 f

• Choosing f larger than optimum has little


effect on delay and reduces # stages
– f=4 is common practice for
– too many stages has negative impact on delay
217
Image taken from: http://bwrcs.eecs.berkeley.edu/Classes/icdesign/ee141_f01/Notes/chapter5.pdf

Example: buffer design


N f tp
1 64 1 64 65

1 8 64 2 8 18

1 4 16 64 3 4 15

1
22.6 64 4 2.8 15.3
2.8 8

218

13
6/8/2018

Example: size the buffers

always
Cext/Cint C1
CL=64C1

219

Example: size the buffers (cont’d)

CL=64C1
C1
buffer 2 is 3.6x larger
than buffer 1

buffer 3 is 8.8x larger


than buffer 1

220

14
6/8/2018

Examples

Delay minimization

221

Two cases for delay minimization


in out
1 2 N
Cg,1 CL

• Known number of gates N, unknown f


– Find optimal f such that F=fN
• Known fopt, unknown N
– Find N with f as close as possible to fopt

222
Image adapted from: Digital Integrated Circuits (2nd Edition) by Rabaey, Chandrakasan, Nikolic

15
6/8/2018

Example: propagation delay tp


1

Cg 1 4 16 CL
CW

223

16

You might also like