logistics - university of michigan

22
1 EECS 427 Lecture 3: CMOS review EECS 427 F09 Lecture 3 1 Reading: 5.4, 6.2 Logistics CAD1 due today at 7pm CAD2 assigned yesterday, due next Monday (9/21) at 7pm HW1 due next Wednesday (9/23) at the beginning of lecture HW1 readings: 3.5, 12.2, 5.4, 6.2.3 HW2 due in 1.5 weeks Email zhengya at eecs.umich.edu, subject line [EECS427 P j tG ]1 il i ld b k d f h Project Group]: 1 email per group, include background of each person: grad/undergrad, area of interest/specialty: circuits, architecture, … Good to mix EE and CE, grad and undergrad, foreign and domestic, male and female EECS 427 F09 Lecture 3 2

Upload: others

Post on 30-May-2022

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Logistics - University of Michigan

1

EECS 427Lecture 3: CMOS review

EECS 427 F09 Lecture 3 1

Reading: 5.4, 6.2

Logistics

• CAD1 due today at 7pm• CAD2 assigned yesterday, due next Monday (9/21) at g y y, y ( )

7pm

• HW1 due next Wednesday (9/23) at the beginning of lecture

• HW1 readings: 3.5, 12.2, 5.4, 6.2.3

• HW2 due in 1.5 weeks– Email zhengya at eecs.umich.edu, subject line [EECS427

P j t G ] 1 il i l d b k d f hProject Group]: 1 email per group, include background of each person: grad/undergrad, area of interest/specialty: circuits, architecture, …

– Good to mix EE and CE, grad and undergrad, foreign and domestic, male and female

EECS 427 F09 Lecture 3 2

Page 2: Logistics - University of Michigan

2

Outline

• Last time: layouts and design rules

• Register design for CAD2• CMOS inverter• DC and dynamic operations• Propagation delay• Static CMOS gates

EECS 427 F09 Lecture 3 3

• Fast gates design techniques

Register Design

Tri-state bufferTri state buffer

Transmission gate

EECS 427 F09 Lecture 3 4

Page 3: Logistics - University of Michigan

3

Register Design

• How it worksWhen C is equal to 0, the master latch is transparent, and the salve latch is closed.

Master latch

Slave latch

EECS 427 F09 Lecture 3 5

Register Design

• How it worksWhen C is equal to 1, the master latch is closed, and the salve latch is

transparent When C switches from 0->1 (rising edge of the clock) the register captures the data .

Master latch

Slave latch

EECS 427 F09 Lecture 3 6

Page 4: Logistics - University of Michigan

4

The CMOS InverterVDD

Vin Vout

EECS 427 F09 Lecture 3 7

CMOS Inverter VTCVout NMOS off

PMOS res

1.5

22.

5

NMOS satPMOS sat

NMOS satPMOS res

EECS 427 F09 Lecture 3 8Vin0.5 1 1.5 2 2.5

0.5

1

NMOS resPMOS off

NMOS resPMOS sat

Page 5: Logistics - University of Michigan

5

Impact of Sizing

2.5

1

1.5

2V

ou

t(V)

Wider PMOS

Wider NMOS

Symmetrical

EECS 427 F09 Lecture 3 9

0 0.5 1 1.5 2 2.50

0.5

Vin

(V)

Impact of Process Variation2.5

1

1.5

2

Vo

ut(V

)

Fast PMOSSlow NMOS

Fast NMOSSlow PMOS

Nominal

EECS 427 F09 Lecture 3 10

0 0.5 1 1.5 2 2.50

0.5

Vin

(V)

Page 6: Logistics - University of Michigan

6

MOS Transistor as a Switch

Discharging a capacitorDischarging a capacitor

C

EECS 427 F09 Lecture 3 11

MOS Transistor as a SwitchTraversed path

ID VGS = VDD

VGS ≥ VTRon D

VDS

VDDVDD /2

GS DD

Rmid

R0

S Don

Lecture 3 12

VDDVDD /2

2

112

2

112

21 )(

)(1)(

1))((

t

t D

DSt

ton

tttoneq dt

tItV

ttdttR

tttRavgR

)()(21

21 tRtRR ononeq

Page 7: Logistics - University of Michigan

7

MOS Transistor as a Switch

VGS ≥ VTRon

ID VGS = VDD

S D

VDS

VDDVDD /2

Rmid

R0

mideq RRR 02

1

Lecture 3 13

21

212

1

DDDSAT

DD

DDDSAT

DDeq VI

VVI

VR

DD

DSAT

DDeq V

IV

R65

143

Gate-Channel CapacitanceG

CGC

G

CGC

G

CGC

S D S D S D

Cut-off Resistive Saturation

CGCB CGCS CGCD

EECS 427 F09 Lecture 3 14

Off/Lin Cgate = Cox·W·Leff

Textbook: page 109

Sat Cgate = (2/3)·Cox·W·Leff ox

oxox t

C

Page 8: Logistics - University of Michigan

8

Gate Overlap CapacitancePolysilicon gatePolysilicon gate

xd xd

L d

Top view

Gate-bulkoverlap

Source

n+

Drain

n+W

xd xd

L d

Top view

Gate-bulkoverlap

Source

n+

Drain

n+W

tox

n+ n+

Cross section

L

Gate oxidetox

n+ n+

Cross section

L

Gate oxide

EECS 427 F09 Lecture 3 15

doxO xCC

Top viewTop view

Off/Lin/Sat CGSO = CGDO = CO·W

Diffusion Capacitance

Side wall

NA+

Bottom

Side wallChannel

Source

Substrate

W

LS

ND

xj

EECS 427 F09 Lecture 3 16

NA

Cdiff = Cbottom + Csw

= Cj · AREA + Cjsw · PERIMETER

Off/Lin/Sat Cdiff = Cj·LS·W + Cjsw·(2LS+W)

Page 9: Logistics - University of Michigan

9

Computing the Capacitances

VDD

VDD

13

Miller effect

M3

M4

M1

M2

Cw Cg3Cdb1

Cg4

Vout2

Cdb2

VinVout

Cgd12

2

3

Reverse

Off Sat (M4)

Lin (M3)

4

No Miller

EECS 427 F09 Lecture 3 17

FanoutVoutVin

CL

SimplifiedModel

biased junction

No Miller effect

Propagation Delay(Approach 1)

VDD

Vout

CLI

tpHL = CL Vswing/2

Iav

CL

kn VDD

~

EECS 427 F09 Lecture 3 18

Vin = VDD

CLIav kn VDD

Page 10: Logistics - University of Michigan

10

Propagation Delay(Approach 2)

VDD

Vout

R

CL

tpHL = f(Ron.CL)

= 0.69 RonCL

Vout

VDD1

ln(0.5)

EECS 427 F09 Lecture 3 19

Vin = VDD

Ron

t

DD

RonCL

0.5

0.36

Transient Response

3 ?

1

1.5

2

2.5

Vo

ut(V

)

tpHL = 0.69 CL Reqn

tpLH = 0.69 CL Reqp

?

tpLHtpHL

EECS 427 F09 Lecture 3 20

0 0.5 1 1.5 2 2.5

x 10-10

-0.5

0

0.5

t (sec)

Page 11: Logistics - University of Michigan

11

Delay as a Function of VDD

5.5

2 5

3

3.5

4

4.5

5

t p(n

orm

aliz

ed

)

EECS 427 F09 Lecture 3 21

0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.41

1.5

2

2.5

VDD

(V)

Device Sizing

3.6

3.8x 10

-11

(for fixed load)

2.6

2.8

3

3.2

3.4

3.6

t p(s

ec)

( )

Self-loading effect:Intrinsic capacitancesdominate

EECS 427 F09 Lecture 3 22

2 4 6 8 10 12 142

2.2

2.4

S

dominate

Page 12: Logistics - University of Michigan

12

PMOS/NMOS Ratio

5x 10

-11

4

4.5

t p(s

ec)

tpLH tpHL

tp = Wp/Wn

EECS 427 F09 Lecture 3 23

1 1.5 2 2.5 3 3.5 4 4.5 53

3.5

Input Rise Time0.35

t pH

L(n

sec

)

0.3

0.25

0.2

EECS 427 F09 Lecture 3 24

0.15

trise (nsec)10.80.60.40.20

tp = tstep(i) + tstep(i-1)

Page 13: Logistics - University of Michigan

13

Static CMOS Circuits

At every point in time (except during the switching transients) each gate output is connected to eithertransients) each gate output is connected to eitherVDD or Vssvia a low-resistive path.

The outputs of the gates assume at all times the value of the Boolean function, implemented by the circuit (ignoring, once again, the transient effects during switching periods).

This is in contrast to the dynamic circuit style which

EECS 427 F09 Lecture 3 25

This is in contrast to the dynamic circuit style, which relies on temporary storage of signal values on the capacitance of high impedance circuit nodes.

Static CMOSVDD

I

F(In1,In2,…InN)

In1

In2

InN

In1

In2

InN

PUN

PDN

PMOS only

NMOS only

EECS 427 F09 Lecture 3 26

PUN and PDN are dual logic networksPUN and PDN functions are complementary

Page 14: Logistics - University of Michigan

14

Threshold DropsVDD

PUNVDD

VS D

VDD 0PDN

0 VDD

CL

0 VDD - VTn

CL

VDD

VDD |VTp|

D SVGS

V

EECS 427 F09 Lecture 3 27

CLVDDCL

S

SD

D

VGS

NAND Gate

EECS 427 F09 Lecture 3 28

PDN: G = AB Conduction to GND

PUN: F = A + B = AB Conduction to VDD

G(In1,In2,In3,…) ≡ F(In1,In2,In3,…)

Page 15: Logistics - University of Michigan

15

NOR Gate

EECS 427 F09 Lecture 3 29

Complex CMOS Gate

B

OUT = D + A • (B + C)

A

D

AC

EECS 427 F09 Lecture 3 30

D

B C

Page 16: Logistics - University of Michigan

16

CMOS Properties

• Full rail-to-rail swing; high noise margins• Logic levels not dependent upon the relativeLogic levels not dependent upon the relative

device sizes; ratioless• Always a path to Vdd or Gnd in steady state;

low output impedance• Extremely high input resistance; nearly zero

steady-state input current

EECS 427 F09 Lecture 3 31

• No direct path steady state between power and ground; no static power dissipation

• Propagation delay function of load capacitance and resistance of transistors

Input Pattern Affects Delay

• Delay is dependent on th tt f i tthe pattern of inputs

• Low to high transition– both inputs go low

• delay is 0.69 Rp/2 CL

– one input goes low• delay is 0.69 Rp CL

CL

B

Rn

A

Rp

B

Rp

R

EECS 427 F09 Lecture 3 32

• High to low transition– both inputs go high

• delay is 0.69 2Rn CL

A

Rn Cint

Page 17: Logistics - University of Michigan

17

Delay Dependence on Input Patterns

3Input Data Delay

0.5

1

1.5

2

2.5 A=B=10

A=1, B=10

A=1 0, B=1

Vol

tage

[V

]

Input Data

Pattern

Delay

(psec)

A=B=01 67

A=1, B=01 64

A= 01, B=1 61

A=B=10 45

EECS 427 F09 Lecture 3 33

-0.5

00 100 200 300 400

time [ps]

A=1, B=10 80

A= 10, B=1 81

NMOS = 0.5m/0.25 mPMOS = 0.75m/0.25 mCL = 100 fF

Transistor Sizing

R R R

CL

B

Rn

A

Rp

B

Rp

R

B

Rp

A

Rp

R R C

Cint

2

2 2 4

4

EECS 427 F09 Lecture 3 34

A

Rn Cint

A

Rn

B

Rn CL21

1

Page 18: Logistics - University of Michigan

18

Transistor Sizing

OUT = D + A • (B + C)

D

A

B

C

4

4

8

8

EECS 427 F09 Lecture 3 35

D

A

B C

1

2

2 2

Transistor Sizing

OUT = D + A • (B + C)

D

A

B

C

6

3

6

6

EECS 427 F09 Lecture 3 36

D

A

B C

1

2

2 2

Page 19: Logistics - University of Michigan

19

Fan-In Considerations

DCBA

C

B

A CL

C3

C2

Distributed RC model(Elmore delay)

tpHL = 0.69 Reqn(C1+2C2+3C3+4CL)

EECS 427 F09 Lecture 3 37

D C1

pHL eqn( 1 2 3 L)

Propagation delay deteriorates rapidly as a function of fan-in – quadraticallyin the worst case.

Fast Gates Design Technique

• Transistor sizing– as long as fan-out capacitance dominates

• Progressive sizing

InN CLMNDistributed RC line

M1 > M2 > M3 > … > MN(the FET closest to the

EECS 427 F09 38

C3

C2

C1In1

In2

In3

M1

M2

M3

(the FET closest to theoutput is the smallest)

Can reduce delay by more than 20%; Be careful: input loading, junction caps, decreasing gains as technology shrinks

Page 20: Logistics - University of Michigan

20

Fast Gates Design Technique

• Transistor ordering

C2In2

In3

M2

M3 CL

C2In2

In1

M2

M3 CL

critical path critical path

charged1

charged1

1

01 charged

discharged

EECS 427 F09 Lecture 3 39

C1In1 M1 C1

In3 M101

charged

delay determined by time to discharge CL, C1 and C2

delay determined by time to discharge CL

1 discharged

Fast Gates Design Technique

• Alternate logic structures

F = ABCDEFGH

EECS 427 F09 Lecture 3 40

Page 21: Logistics - University of Michigan

21

Fast Gates Design Technique

• Isolating fan-in from fan-out using buffer insertion

CLCL

EECS 427 F09 Lecture 3 41

Fast Gates Design Technique

• Reducing the voltage swing

– linear reduction in delay– also reduces power consumption

B t th f ll i t i h l !

tpHL = 0.69 (3/4 (CL VDD)/ IDSATn )

= 0.69 (3/4 (CL Vswing)/ IDSATn )

EECS 427 F09 Lecture 3 42

• But the following gate is much slower!• Or requires use of “sense amplifiers” on the

receiving end to restore the signal level (memory design)

Page 22: Logistics - University of Michigan

22

Summary

• CMOS is the dominant circuit family due to:N i i– No static power consumption

– Ease of design

– Robust to variations and noise

• CAD2 due next Monday at 7 pm

• HW1 due next Wednesday at 1:30 pm

EECS 427 F09 Lecture 3 43

• Form a group. Email zhengya at eecs.umich.edu