54
EEL7312 – INE5442 Digital Integrated Circuits 1 Digital Integrated Circuits Chapter 5 - Interconnections

Digital Integrated Circuits - Departamento de …guntzel/ine5442/interconnections.pdf · EEL7312 – INE5442 Digital Integrated Circuits 12 Circuit Simulation - 1 Why using circuit

  • Upload
    hathien

  • View
    221

  • Download
    0

Embed Size (px)

Citation preview

EEL7312 – INE5442

Digital Integrated Circuits

1

Digital Integrated Circuits

Chapter 5 - Interconnections

EEL7312 – INE5442

Digital Integrated Circuits

2

Contents

Introduction

Resistance

Capacitance

RC delay

Inductance

Interconnection modeling

Scaling effects on interconnection

EEL7312 – INE5442

Digital Integrated Circuits

3

Introduction - 1

Why are on-chip interconnects important? As technology scales to deep submicron:

Increased contribution to propagation delay Increased contribution on energy dissipation

Introduces extra noise, affects reliability

Trend toward higher integration levels partially driven

by faster, denser, and more reliable on-chip than off-chip interconnects.

Interconnect modeling: resistors, capacitors, and

inductors.

EEL7312 – INE5442

Digital Integrated Circuits

4

Introduction - 2

physical

Source: Rabaey

schematics

transmitters receivers

EEL7312 – INE5442

Digital Integrated Circuits

5

Introduction - 3

Wire Models

All-inclusive modelCapacitance-only

Source: Rabaey

EEL7312 – INE5442

Digital Integrated Circuits

6

Introduction - 4

Source: Weste

EEL7312 – INE5442

Digital Integrated Circuits

7

Introduction - 5

130 nm CMOS technology (Intel)

Transistor

Via

M1

M2

M3

M4

M5

M6

Isolation

EEL7312 – INE5442

Digital Integrated Circuits

8

Resistance - 1

L LR R

H W W

ρ= =

R

: sheet resistance

R1 R2

W

L

H

I

Material

property

Height

Source: Rabaey

Defined by

manufacturer

Defined by

designer

(sometimes)

Ohm’s law

/I V R=

RI

+ V -

EEL7312 – INE5442

Digital Integrated Circuits

9

Resistance - 2

Source: Rabaey

EEL7312 – INE5442

Digital Integrated Circuits

10

Ex: Poly-resistor

L

W

Polysilicon

P substrate

Metal Metal

2 H

LR R R

W= +

R

: sheet resistance

HR : head resistance

Top view

Resistance - 3

EEL7312 – INE5442

Digital Integrated Circuits

11

Resistance - 4

Source: Rabaey

Example: Calculate the approximate resistance of a 1 µm-wide, 1 mm-long wire

of (a) polysilicon; (b) aluminum. Use the data of the above table.

Sheet resistance values for a typical 0.25 µm CMOS process

EEL7312 – INE5442

Digital Integrated Circuits

12

Circuit Simulation - 1

Why using circuit simulators?

Designs can be quickly evaluated without (sometimes very expensive)

fabrication.

After design has been evaluated you can prototype it before mass

production.

A circuit simulator computes the response of the circuit to a particular stimulus.

The simulator formulates the circuit equations and then numerically solves them.

Types of analyses:

DC/DC sweep: Both stimuli and responses do not vary with time

Transient: Responses vary with time

AC/Noise: also called small-signal analysis, it computes the sinusoidal steady-state response

Source: Kundert

EEL7312 – INE5442

Digital Integrated Circuits

13

Circuit Simulation - 2

What are the input data?

Device Type (R, C, L, current sources, voltage sources, diodes,

transistors)

Device models/parameters/ dimensions

How devices are connected

Source: Kundert

Some circuit simulators:

SPICE, PSPICE, HSPICE, Spectre, Smash, SPiceOpus,….

EEL7312 – INE5442

Digital Integrated Circuits

14

Simulation 4.1

Use SpiceOpus to determine the (dc) I-V characteristic of a 1 kΩ resistor.

resistortest

* this is resistortest.cir file

v0 1 0 dc 10V

r1 1 0 1k

.end

r1

I

V0+-

Node 1

Node 0

SpiceOpus (c) 1 -> source resistortest.cir SpiceOpus (c) 2 -> dc v0 -1V 1V 2mV SpiceOpus (c) 3 -> setplot dc1 SpiceOpus (c) 4 -> plot i(v0) xlabel v(1) ylabel current[A] SpiceOpus (c) 5 -> plot -1000*i(v0) xlabel v(1)[V] ylabel current[mA]

V=RI

RI

+ V -

EEL7312 – INE5442

Digital Integrated Circuits

15

Simulation 4.1

r1= 1 kΩ

I

V0+-

Node 1

Node 0

SpiceOpus (c) 5 -> source currentdivider.cir

SpiceOpus (c) 6 -> dc v0 -1 1 5m

SpiceOpus (c) 7 -> setplot dc

SpiceOpus (c) 8 -> plot 1000*i(v1) xlabel v(3)[V] ylabel current[mA]

EEL7312 – INE5442

Digital Integrated Circuits

16

Exercise 4.1 Use SpiceOpus to determine the (dc) I-V0

transfer characteristic of the circuit given below.

r1= r2 =1 kΩ

I

V0+- r1 r2

r3

r3= 0.5 kΩ1V ≥ V0≥ -1V

EEL7312 – INE5442

Digital Integrated Circuits

17

Capacitance - 1

Source: wikipedia

The parallel plate capacitor

Charge separation in a parallel-plate capacitor

causes an internal electric field. A polarized

dielectric spacer (orange) reduces the electric

field and increases the capacitance.

Capacitance: is a measure of the charge

stored on each plate for a given voltage

such that Q=CV

The electric field (force) E between the

plates of a parallel plate capacitor is

uniform and given by E=V/d

EEL7312 – INE5442

Digital Integrated Circuits

18

Capacitance - 2

Source: Rabaey

Dielectric

Substrate

L

W

H

tdi

Electrical-field lines

Current flow

WLt

cdi

diint

ε=

2 fF/µmoxox

ox

Ct

ε=

Defined by foundry

EEL7312 – INE5442

Digital Integrated Circuits

19

Capacitance - 3

Source: Rabaey

material r oε ε ε=128.85 10 F/m

oε −= ⋅

: permittivity of free spaceo

ε

EEL7312 – INE5442

Digital Integrated Circuits

20

Capacitance - Problem

material r oε ε ε=128.85 10 F/m

oε −= ⋅

: permittivity of free spaceo

ε

The distance between the plates of a parallel-plate capacitor

is 10-4 mm. The area of the plates is 1 cm2. Find:

a)The capacitance if the space between the plates is filled

with air

b)The capacitance if the dielectric is SiO2

c)The electric field for 1V voltage difference between the

plates

d)The electric charge Q stored in the SiO2 capacitor for 1V

and the charge density in C/cm2. Give the number of

missing/excess electrons in each plate

Answer

a) C=8.85⋅10-9 F = 8.85 nF

b) εrSiO2= 3.9 ; C=3.9⋅ 8.85=34.5 nF

c) E=107 V/m

d) Q= 34.5 nC; Q/A= 34.5/1=34.5 nC/cm2. N=±Q/q;

N=±21.6 ⋅1010 electrons

EEL7312 – INE5442

Digital Integrated Circuits

21

Capacitance - 4

Source: MOSIS

Fabrication

process

(CMOS)

Gate oxide

thickness

(nm)

Capacitance

/ area

(fF/µm2)

AMIS 1.5

µm

32 1.1

IBM 0.25

µm

6.3 5.5

IBM 0.13

µm

3.2 11

Source: Intel Tech. Journal

EEL7312 – INE5442

Digital Integrated Circuits

22

Capacitance - 5

for constant capacitance

; / ( ) /Q CV I dQ dt d CV dt= = =

/I CdV dt=

+

V

-

I For constant V→ I=0, i.e. a capacitor behaves as an open circuit at

dc.

Capacitors are energy-storage (memory) devices used in filters,

oscillators, power sources, DRAM,…..

Ideal capacitors are not dissipative (and not noisy) but charging and

discharging them causes heating through dissipative devices

connected to the capacitors.

EEL7312 – INE5442

Digital Integrated Circuits

23

The RC circuit - 1

KCL / /C RI CdV dt V R= =

Assume that VS=0 for t<0,

VS=A for t≥0 (and VC(0)=0).

S C RV V V= +KVL

/S C CV RCdV dt V= +

( )1 exp / 0;

0 0

C

C

V A t t

V t

τ= − − ≥

= <

( )/ exp / / 0CI CdV dt A t R tτ= = − ≥

+

VC

-

RI

+ VR -

VS+- C

td=τ ln2≅0.69 τRCτ =

td

50%

EEL7312 – INE5442

Digital Integrated Circuits

24

The RC circuit - 2

Assume that VS=0 for t<0,

VS=A for t≥0 (and VC(0)=0).

( )1 exp /C

V A t τ= − − ( )exp / / 0I A t R tτ= − ≥

The power dissipation p (electric power converted into

heat) in the resistor is

( )2 2 exp 2 / /p RI A t Rτ= = −

The energy converted into heat in the resistor is2 2

0 0

2exp

2R

A t CAE pdt dt

R τ

∞ ∞ = = − =

∫ ∫The energy stored in the capacitor (for t>>τ)

2 2

2 2

CC

CV CAE = =

Exercise: (a) Using the energy conservation principle calculate the energy delivered by

the source. (b) Calculate the energy ES delivered by the source using the formula below

0

S SE V Idt

= ∫

+

VC

-

RI

+ VR -

VS+- C

EEL7312 – INE5442

Digital Integrated Circuits

25

Simulation 4.2

+

VC

-

RI

+ VR -

VS+-

1 2

0

C

RC1

* this is RC1.cir file

v0 1 0 dc 0 pulse 0 1V 0 10ps 10ps 10ns 20ns

R 1 2 1k

C 2 0 1p

.end

EEL7312 – INE5442

Digital Integrated Circuits

26

pulse Vlow Vhigh delay rise fall width period

pulse

v0 1 0 dc 0 pulse 0 1V 0 10ps 10ps 10ns 20ns

Vlow=0

Vhigh=1

delay rise width fall

period0 10ps 10ps10ns

20ns

EEL7312 – INE5442

Digital Integrated Circuits

27

Exercise 4.2 Run SpiceOpus to determine the voltages at the intermediate nodes 2 and 3 for the stimulus of simulation 4.2

R/2I

VS+-

1 2

0C/2

R/2

3

C/2

R= 1 kΩ

C= 1 pF

SpiceOpus (c) 1 -> source RC2.cir SpiceOpus (c) 2 -> tran 0.1ns 20ns

SpiceOpus (c) 3 -> setplot

new New plot

Current tran1RC2 (Transient Analysis)

const Constant values (constants) SpiceOpus (c) 4 -> plot v(1) v(2) v(3) xlabel time[s] ylabel

Outputs[V]

V(3)V(2)

EEL7312 – INE5442

Digital Integrated Circuits

28

Comparison between exercises 4.1 and 4.2

R/2I

VS+-

1 2

0C/2

R/2

3

C/2

R= 1 kΩ

C= 1 pF

RI

VS+-

1 4

0

C

V(4)V(3)

EEL7312 – INE5442

Digital Integrated Circuits

29

Capacitance - 6

Fringing Capacitance

capacitance/unit length

w=W-H/2

W - H/2H

+

(a)

(b)

tdi

substrate

W

H

thick oxide

Source: Rabaey

cfringe

cfringe

cpp

cpp

EEL7312 – INE5442

Digital Integrated Circuits

30

Capacitance - 7

Interwire Capacitance

fringing parallel

Source: Rabaey

Crosstalk: a signal can affect

another nearby signal.

Substrate noise coupling

EEL7312 – INE5442

Digital Integrated Circuits

31

Capacitance - 8

Source: Rabaey

Wiring Capacitances (0.25 µm CMOS)

aF/µm2

aF/µm

EEL7312 – INE5442

Digital Integrated Circuits

32Source: Rabaey

Estimate the capacitance of the wires (with the substrate) specified below:

1. Polysilicon, W= 0.25µm, L=1 mm; 2. Polysilicon, W= 0.25µm, L=10 mm;

3. Metal 1, W= 0.25µm, L=1 mm; 4. Metal 1, W= 0.25µm, L=10 mm.

Exercise 4.3

In each case, calculate the delay time assuming a lumped RC model

for the wire and the capacitance with the substrate. Assume that the

sheet resistances for polysilicon (with silicide) and metal 1 are 5 Ω and

0.1 Ω, respectively.

substrate

W

H

thick oxide

L

; fringePP

PP fringe

CCC WL C L

area length

= =

wire PP fringeC C C= + wire

LR R

W=

0.69d wire wire

t R C≅

288 aF/µm 54 aF/µmfringePP

CC

area length= =

EEL7312 – INE5442

Digital Integrated Circuits

33

1. Cwire=76 fF, Rwire= 20 kΩ, RwireCwire= 1520 ps, td=0.69RwireCwire=1050 ps

Exercise 4.3 - Answer

2. Cwire=760 fF, Rwire= 200 kΩ, RwireCwire= 152 ns, td=0.69RwireCwire=105 ns

3. Cwire=47.5 fF, Rwire= 400 Ω, RwireCwire= 19 ps, td=0.69RwireCwire=13 ps

4. Cwire=475 fF, Rwire= 4 kΩ, RwireCwire= 1.9 ns, td=0.69RwireCwire=1.3 ns

Note that the delay time increases proportionally with the square of the wire length. Why?

So far we have considered that the distributed RC line can be represented by a lumped RC model (pessimistic view) and that the drive signal is a step supplied by an ideal voltage source (optimistic view).

EEL7312 – INE5442

Digital Integrated Circuits

34

RC delay - 1

Vout

Driver

cwire

Vin

Clumped

RdriverVout

Source: Rabaey

Influence of the output resistance of the driver

Rwire<< Rdriver

Example: Rdriver =100 kΩ,

and 1-µm-wide, 10-mm-long Al1 wire. What’s tpd?

230 aF/µm 40 aF/µmfringePP

CC

area length= =

0.3 pF; 0.4 pFfringePP

PP fringe

CCC WL C L

area length

= = = =

0.7 pFwire PP fringeC C C= + =

0.69d driver wire

t R C≅ 50 nsd

t ≅

What’s the approximate maximum operating

frequency of the input such that the output can

detect the correct value of the input?

EEL7312 – INE5442

Digital Integrated Circuits

35

RC delay – 2: The Elmore delay -1

Sources: Rabaey & *W. C. Elmore, “The transient response

of damped linear networks with particular regard to wideband amplifiers,” J. Applied Physics, vol. 19, Jan 1948

Elmore delay model *– method to determine the approximate delay time in an RC

network; it avoids running costly simulations for calculation of delay time. Useful for

determining delays in transmission lines, gates, clock distribution networks,…

EEL7312 – INE5442

Digital Integrated Circuits

36Source: Rabaey

path s→i Rii=R1+R3+Ri

path s→1 Ri1=R1

path s→2 Ri2=R1

path s→3 Ri3=R1+R3

path s→4 Ri4=R1+R3

RC delay – 3: The Elmore delay -2

( )

( )( )

1 1 2 2 3 3 4 4

1 3

1 1

1 2

1 3 3

1 3 4

Di i i i i ii i

i i

R C R C R C R C R C

R R R C

R C

R C

R R C

R R C

τ = + + + + =

+ + +

+

+

+ +

+ +

EEL7312 – INE5442

Digital Integrated Circuits

37

RC delay – 4: The Elmore delay - 3

Let R, C, and l be the total line resistance, capacitance, and length.

Application of the Elmore delay formula to a (RC) wire.

/ ; / ; /r R l c C l L l N= = ∆ =

( ) ( ) ( )

( )

2

1

2 2

1 2 ....

1 1 /

2 2

N

Dout

i

ir L c L rc L N

N Nrc l N N rcl

N

τ=

= ∆ ∆ = ∆ + + + =

+ +=

∑2

2 1lim

2 2 2Dout

N

N rcl RCrcl

→∞

+= = =

The delay of a wire is proportional

to the square of its length. Note: The Elmore formula applied to the

RC lumped model gives τDout=RC

Source: Rabaey

EEL7312 – INE5442

Digital Integrated Circuits

38

RC delay – 5: The Elmore delay - 4

Example 4.8 of Rabaey’s book: 10-cm-long, 1- µm-wide Al1 wire

for which r=0.075 Ω/ µm, c= 110 aF/µm.

( )22 5/ 2 0.075 /µm 110aF/µm 10 µm / 2 41.3 ns

Doutrclτ = = Ω ⋅ ⋅ =

Note: The Elmore delay is, in general, not equal to the delay time. For

a distributed RC network, the Elmore delay τD = 0.5 RC whereas the delay time td = 0.38 RC

Source: Rabaey

EEL7312 – INE5442

Digital Integrated Circuits

39

RC delay – 6Example 4.8 of Rabaey’s book: 10-cm-

long, 1- µm-wide Al1 wire for which

r=0.075 Ω/ µm, c= 110 aF/µm.

Distributed RC line 1

* this is DistributedRCline.cir file

v0 1 0 dc 0 pulse 0 1V 0 10ps 10ps 200ns 400ns

URC1 1 2 0 MURC L=100m

.model MURC URC rperl=75k cperl=110p

.end

SpiceOpus (c) 7 -> source DistributedRCline.cir

SpiceOpus (c) 8 -> tran 1ns 200ns

SpiceOpus (c) 9 -> setplot

new New plot

Current tran2 Distributed RC line 1 (Transient Analysis)

SpiceOpus (c) 10 -> setplot tran2

SpiceOpus (c) 11 -> plot v(2) xlabel time ylabel Vout

lumped

distributed

EEL7312 – INE5442

Digital Integrated Circuits

40

RC delay – 7

Diffusion equation

Source: Rabaey

EEL7312 – INE5442

Digital Integrated Circuits

41

RC delay – 8

Step-response of RC wire as a function of time and space

Source: Rabaey

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

0.5

1

1.5

2

2.5

time (nsec)

vo

lta

ge

(V

)

x= L/10

x = L/4

x = L/2

x= L

EEL7312 – INE5442

Digital Integrated Circuits

42

RC delay – 9

Voltage range Lumped RC network Distributed RC network

0→50% (tp) 0.69 RC 0.38 RC

0→63% (τ) RC 0.5 RC

10→90% (tr) 2.2 RC 0.9 RC

Source: Rabaey

EEL7312 – INE5442

Digital Integrated Circuits

43

RC delay – 10

Source: Rabaey

Vout

Driver

cwire

VinC

Rdriver

Vout

L

rw, cw, L

When are the effects of the

wire delay important?

Assume that the driver delay

is tpgate. The wire delay is 2

0.38 0.38pwire w wt RC r c L= =

The wire delay is important when

tpwire≅tpgate or, equivalently

0.38

pgate

crit

w w

tL

r c=

EEL7312 – INE5442

Digital Integrated Circuits

44

RC delay – 11Example 4.8 of Rabaey’s book: 10-cm-

long, 1- µm-wide Al1 wire for which

r=0.075 Ω/ µm, c= 110 aF/µm.

Distributed RC line 2

* this is DistributedRCline2.cir

*file

* the rise time is of the order of the

*RC time constant

v0 1 0 dc 0 pulse 0 1V 0 50ns 50ns

+200ns 500ns

URC1 1 2 0 MURC L=100m

.model MURC URC K=2

+fmax=20G rperl=75k cperl=110p

.end

Response to pulse

rise time=0

Response to pulse

rise time=50 ns

Note that the internal resistance of the

voltage source is zero in this example

What if the rise time becomes much higher than RC?

EEL7312 – INE5442

Digital Integrated Circuits

45

RC delay – 12Example 4.8 of Rabaey’s book: 10-cm-

long, 1- µm-wide Al1 wire for which

r=0.075 Ω/ µm, c= 110 aF/µm.

What if the rise time becomes much higher than RC?

EEL7312 – INE5442

Digital Integrated Circuits

46

RC delay – 13

Source: Weste&Harris

EEL7312 – INE5442

Digital Integrated Circuits

47

RC delay – 14

Source: Rabaey

Design Rules of Thumb

rc delays should only be considered when tpRC >> tpgate of the driving gate

Lcrit >> √ tpgate/0.38rc

rc delays should only be considered when the rise (fall) time at the line input is smaller than RC, the rise (fall) time of the line

trise < RC when not met, the change in the signal is slower than

the propagation delay of the wire

EEL7312 – INE5442

Digital Integrated Circuits

48

I

+ VL -

Inductance - 1

/LV LdI dt=2 / 2LE LI=

Inductive effects

important for power grids (high current), clock networks

(high speed), and wide busses (low resistance/unit length);

may cause ringing/overshoot effects, reflection of signals,

inductive coupling between lines (crosstalk), and switching

noise in power lines

Clock trees and power/ground grid need to be designed carefully to avoid large clock skew, signal inductive coupling and ground bounce

EEL7312 – INE5442

Digital Integrated Circuits

49

Inductance - 2

Inductance of a wire depends on its geometry and surrounding dielectric

Extracting the inductance is in general a 3-D problem and is extremely time-consuming for complex geometries

Inductance depends on the entire current loop; it is impractical to extract the inductance from a chip layout

Source: Rabaey,

Weste&Harris

EEL7312 – INE5442

Digital Integrated Circuits

50

Inductance - 3

The Wave Equation

V in Voutr

c

r r x

c

r

c c

l l l l

The Transmission Line

Source: Rabaey

When r=0 →→→→ signal travels at speed of light, which is smaller than speed of light in vacuum (300 mm/ns). In the real case, currents return in distant power lines and

increase inductance thus reducing signal velocity.

When l=0 →→→→ rc wire (diffusion equation)

EEL7312 – INE5442

Digital Integrated Circuits

51

Inductance - 4

IEEE JSSC, vol. 37, no. 6, June 2002

High-frequency characterization of on-chip digital interconnects

Kleveland, B.; Xiaoning Qi; Madden, L.; Furusawa, T.; Dutton, R.W.;

Horowitz, M.A.; Wong, S.S.

Page(s): 716-725

EEL7312 – INE5442

Digital Integrated Circuits

52

Inductance - 5

Source: Qi, CICC 2000

EEL7312 – INE5442

Digital Integrated Circuits

53

Crosstalk is the coupling of energy from one line to another via:

Mutual capacitance (electric field)

Mutual inductance (magnetic field)

Mutual Capacitance, Cm Mutual Inductance, Lm

Source: Intel

Zs

Zo

Zo

Zo

Zs

Zo

Zo

Zo

Cm

Lm

near

far

near

far

Crosstalk

EEL7312 – INE5442

Digital Integrated Circuits

54

Scaling effects on interconnection

Rabaey’s book pages 170-174

Weste & Harris, pages 249-250

Interconnect Scaling, Prof. Krishna Saraswat,

On line available

http://www.stanford.edu/class/ee311/NOTES/InterconnectScalingSlides.pdf