EE3230 L5 Circuit Characterization II

Download as pdf or txt
Download as pdf or txt
You are on page 1of 44

EE3230 Lecture 5:

Circuit Characterization and


Performance Estimation II
Ping-Hsuan Hsieh ( )
Delta Building R908
EXT 42590
phsieh@ee.nthu.edu.tw

EE3230 Ping-Hsuan Hsieh


Outline
• Delay estimation
• Logical effort and transistor sizing
• Power dissipation
• Interconnect
• Wire engineering
• Design margin
• Reliability
• Scaling

EE3230 Ping-Hsuan Hsieh 2


Power and Energy
• Power is drawn from a voltage source attached to
the VDD pin(s) of a chip

• Instantaneous Power:

• Energy:

• Average Power:

EE3230 Ping-Hsuan Hsieh 3


Static and Dynamic Power Dissipation

• Static power dissipation


– Sub-threshold conduction through OFF transistors
– Tunneling current through gate oxide
– Leakage through reverse-biased diodes
– Contention current in ratioed circuits
• Dynamic power dissipation
– Charging and discharging of load capacitance
– “Short-circuit” current while both PMOS and NMOS
networks are partially ON

EE3230 Ping-Hsuan Hsieh 4


Dynamic Power (I)
• Dynamic power is required to charge and discharge
load capacitances when transistors switch
• One cycle involves a rising and falling output.
• On rising output, charge Q = CVDD is required
• On falling output, charge is dumped to GND
• This repeats fsw times per second

EE3230 Ping-Hsuan Hsieh 5


Dynamic Power (II)

EE3230 Ping-Hsuan Hsieh 6


Activity Factor
• Suppose the system clock frequency = f
• Let fsw = αf, where α = activity factor
– If the signal is a clock, α = 1
– If the signal switches once per cycle, α = ½
– Dynamic gates: Switch either 0 or 1 times per cycle, a = ½
– Static gates: Depends on design, typically α = 0.1

• Dynamic power:

EE3230 Ping-Hsuan Hsieh 7


Short-Circuit Current
• When transistors switch, both NMOS and PMOS
networks may be momentarily ON at once
• Leads to a blip of short-circuit current
• < 10% of dynamic power if rise/fall times are
comparable for input and output (well-controlled)

EE3230 Ping-Hsuan Hsieh 8


Example
• 200M transistor chip
– 20M logic transistors
• Average width: 12 λ
– 180M memory transistors
• Average width: 4 λ
– 1.2-V 100-nm process (λ = 0.5* feature size = 50nm)
– Cg = 2 fF/μm

EE3230 Ping-Hsuan Hsieh 9


Dynamic Power Consumption
• Static CMOS logic gates: activity factor = 0.1
• Memory arrays: activity factor = 0.05 (many banks
and partially activated at a time!)
• Estimate dynamic power consumption per MHz.
– Neglect wire capacitance and short-circuit current

EE3230 Ping-Hsuan Hsieh 10


Static Power Consumption
• Static power is consumed even when chip is
quiescent.
– Ratioed circuits burn power in fight with ON transistors
– Leakage draws power from nominally OFF devices

EE3230 Ping-Hsuan Hsieh 11


Ratioed Example
• The chip contains a 32 word x 48 bit ROM
– Uses 1:32 pseudo-nMOS decoder and bit-line pull-ups
– In average, one wordline and 24 bitlines are high
• Find static power drawn by the ROM
– β = 75 μA/V2, VDD = 1.8 V
– Vtp = –0.4V

EE3230 Ping-Hsuan Hsieh 12


Leakage Example (I)
• The process has two threshold voltages and two
oxide thicknesses.
• Subthreshold leakage:
– 20 nA/µm for low Vth devices
– 0.02 nA/µm for high Vth devices
• Gate leakage:
– 3 nA/μm for thin oxide
– 0.002 nA/μm for thick oxide
• Memories use low-leakage transistors everywhere
• Gates use low-leakage transistors on 80% of logic

EE3230 Ping-Hsuan Hsieh 13


Leakage Example (II)
• Estimate static power:
– High leakage:
– Low leakage:

• Withow leakage devices, Pstatic = 749 mW (!)

EE3230 Ping-Hsuan Hsieh 14


Low Power Design
• To rduce dynamic power
– α: clock gating, sleep mode
– C: small transistors (esp. on clock), short wires
– VDD: lowest suitable voltage
– f: lowest suitable frequency
• To reduce static power
– Selectively use ratioed circuits
– Selectively use low Vth devices
– Leakage reduction:
stacked devices, body bias, low temperature

EE3230 Ping-Hsuan Hsieh 15


Reduce Static Power
• Leakage stack effect • MTCMOS:
multiple threshold CMOS

• Body bias

EE3230 Ping-Hsuan Hsieh 16


Outline
• Delay estimation
• Logical effort and transistor sizing
• Power dissipation
• Interconnect
• Wire engineering
• Design margin
• Reliability
• Scaling

EE3230 Ping-Hsuan Hsieh 17


Interconnect
• Chips are mostly made of wires called interconnect
– In stick diagrams, wires determine size
– Transistors are little things under the wires
– Many layers of wires
• Wires are as important as transistors
– Speed
– Power
– Noise
• Alternating layers run orthogonally

EE3230 Ping-Hsuan Hsieh 18


Wire Geometry
• Pitch = w + s
• Aspect ratio: AR = t/w
– Old processes had AR << 1
– Modern processes have AR » 2
• Pack in many skinny wires
w s

EE3230 Ping-Hsuan Hsieh 19


Layer Stack
• AMI 0.6-μm process has 3 metal layers
• Modern processes use 6-10+ metal layers
• Example: Intel 180 nm process
• M1: thin, narrow (< 3λ) Layer T (nm) W (nm) S (nm) AR

– High density cells 6 1720 860 860 2.0

• M2-M4: thicker 1000

5 1600 800 800 2.0


– For longer wires
1000

• M5-M6: thickest 4 1080 540 540 2.0


700
– For VDD, GND, clk 3 700
700
320 320 2.2

2 700 320 320 2.2


700
1 480 250 250 1.9
800
Substrate

EE3230 Ping-Hsuan Hsieh 20


Wire Resistance
• ρ = resistivity (Ω*m)

• Ro = sheet resistance (Ω/o)


– o is a dimensionless unit(!) w w

• Count number of squares l

– R = Ro * (# of squares) w

l l

t t

1 Rectangular Block 4 Rectangular Blocks


R = R (L/W) W R = R (2L/2W) W
= R (L/W) W

EE3230 Ping-Hsuan Hsieh 21


Choice of Metals
• Until 180 nm, most wires were aluminum
• Modern processes often use copper
– Cu atoms diffuse into silicon and damage FETs
– Must be surrounded by a diffusion barrier

Metal Bulk resistivity (μΩ*cm)


Silver (Ag) 1.6
Copper (Cu) 1.7
Gold (Au) 2.2
Aluminum (Al) 2.8
Tungsten (W) 5.3
Molybdenum (Mo) 5.3

EE3230 Ping-Hsuan Hsieh 22


Sheet Resistance
• Typical sheet resistances in 180-nm process

Layer Sheet Resistance (Ω/o)


Diffusion (silicided) 3-10
Diffusion (no silicide) 50-200
Polysilicon (silicided) 3-10
Polysilicon (no silicide) 50-400
Metal1 0.08
Metal2 0.05
Metal3 0.05
Metal4 0.03
Metal5 0.02
Metal6 0.02

EE3230 Ping-Hsuan Hsieh 23


Contact Resistance
• Contacts and vias also have 2-20 Ω
• Use many contacts for lower R
– Many small contacts for current crowding around periphery

EE3230 Ping-Hsuan Hsieh 24


Wire Capacitance
• Wire has capacitance per unit length
– To neighbors
– To layers above and below
• Ctotal = Ctop + Cbot + 2Cadj

s w

layer n+1

h2 Ctop

t layer n
Cadj
h1 Cbot
layer n-1

EE3230 Ping-Hsuan Hsieh 25


Capacitance Trend
• Parallel plate equation: C = εA/d
– Wires are not parallel plates, but obey trends
– Increasing area (W, t) increases capacitance
– Increasing distance (s, h) decreases capacitance
• Dielectric constant
– ε = kε0
– ε0 = 8.85 x 10-14 F/cm
– k = 3.9 for SiO2
• Processes are starting to use low-k dielectrics
– k » 3 (or less) as dielectrics use air pockets

EE3230 Ping-Hsuan Hsieh 26


M2 Capacitance Data
• Typical wires have ~ 0.2 fF/μm
– Compare to 2 fF/μm for gate capacitance
400

350

300
M1, M3 planes
s = 320
250 s = 480
s = 640
Ctotal (aF/µm)

s=

8
200
Isolated

150 s = 320
s = 480
s = 640
100
s=

8
50

0
0 500 1000 1500 2000
w (nm)

EE3230 Ping-Hsuan Hsieh 27


Diffusion and Polysilicon
• Diffusion capacitance is very high (about 2 fF/μm)
– Comparable to gate capacitance
– Diffusion also has high resistance
– Avoid using diffusion runners for wires!
• Polysilicon has lower C but high R
– Use for transistor gates
– Occasionally for very short wires between gates

EE3230 Ping-Hsuan Hsieh 28


Lumped Element Models
• Wires are a distributed system
– Approximate with lumped element models

N segments
R R/N R/N R/N R/N
C C/N C/N C/N C/N

R R R/2 R/2

C C/2 C/2 C

L-model p-model T-model

• 3-segment π-model is accurate to 3% in simulation


• L-model needs 100 segments for same accuracy!
• Use single segment π-model for Elmore delay
EE3230 Ping-Hsuan Hsieh 29
Example
• M2 wire in 180-nm process
– 5-mm long
– 0.32-μm wide
• Construct a 3-segment π-model
– Ro = 0.05 Ω/o àR=
– Cpermicron = 0.2 fF/μm àC=

EE3230 Ping-Hsuan Hsieh 30


Wire RC Delay
• Estimate the delay of a 10x inverter driving a 2x
inverter at the end of the 5-mm wire from previous
example
– Effective R = 2.5 kΩ/μm for gates, C = 2 fF/μm
– Unit inverter: 4λ = 0.36 μm nMOS, 8λ = 0.72 μm pMOS

EE3230 Ping-Hsuan Hsieh 31


Crosstalk
• Capacitor do not change voltage instantaneously
• A wire has high capacitance to its neighbor
– When the neighbor (aggressor) switches from 1à0 or
0à1, the wire (victim) tends to switch as well
– Called capacitive coupling or crosstalk
• Impacts
– Cause noise on non-switching wires
– Increase delay on switching wires

EE3230 Ping-Hsuan Hsieh 32


Crosstalk Delay
• Assume layers above and below in average are quiet
– Second terminal of capacitor can be ignored
– Modeled as Cgnd = Ctop + Cbot
• Effective Cadj depends on behavior of neighbors
– Miller Coupling Factor (MCF)

B DV Ceff(A) MCF
Constant VDD Cgnd + Cadj 1 A B
Cadj
Switching with A 0 Cgnd 0 Cgnd Cgnd
Switching opposite A 2VDD Cgnd + 2 Cadj 2

EE3230 Ping-Hsuan Hsieh 33


Crosstalk Noise (Floating Victims)
• Crosstalk causes noise on non-switching wires
• If victim is floating
– modeled as capacitive voltage divider

Aggressor

DVaggressor
Cadj
Victim
Cgnd-v DVvictim

EE3230 Ping-Hsuan Hsieh 34


Crosstalk Noise (Driven Victims)
• Usually victim is driven by a gate that fights noise
– Noise depends on relative resistances
– Assume victim driver in linear region and aggressor driver
in saturation (considering inverter operation)
– With equal sizes, Raggressor = 2-4 x Rvictim

Raggressor
Aggressor
Cgnd-a
DVaggressor
Cadj
Rvictim Victim
Cgnd-v DVvictim

EE3230 Ping-Hsuan Hsieh 35


Coupling Waveforms
• Simulated coupling for Cadj = Cgnd

Aggressor
1.8

1.5

1.2

Victim (undriven): 50%


0.9

0.6
Victim (half size driver): 16%

Victim (equal size driver): 8%


0.3 Victim (double size driver): 4%

0
0 200 400 600 800 1000 1200 1400 1800 2000

t(ps)

EE3230 Ping-Hsuan Hsieh 36


Noise Implications
• So what if we have noise?
• If the noise is less than the noise margin, nothing
happens
• Static CMOS logic will eventually settle to correct
output even if disturbed by large noise spikes
– But glitches cause extra delay
– Also cause extra power from false transitions
• Dynamic logic never recovers from glitches
• Memories and other sensitive circuits also can
produce wrong outputs

EE3230 Ping-Hsuan Hsieh 37


Outline
• Delay estimation
• Logical effort and transistor sizing
• Power dissipation
• Interconnect
• Wire engineering
• Design margin
• Reliability
• Scaling

EE3230 Ping-Hsuan Hsieh 38


Wire Engineering
• Goal: to achieve delay, area, and power goals with
acceptable noise
• Degrees of freedom:

EE3230 Ping-Hsuan Hsieh 39


Wire Engineering
• Goal: to achieve delay, area, and power goals with
acceptable noise
• Degrees of freedom:
– Width 2.0 0.8
1.8 0.7
– Spacing

Coupling:2Cadj / (2C adj+Cgnd)


1.6
0.6
1.4 WireSpacing
0.5 (nm)
Delay (ns):RC/2

1.2
320
1.0 0.4 480
640
0.8 0.3
0.6
0.2
0.4
0.1
0.2
0 0
0 500 1000 1500 2000 0 500 1000 1500 2000
Pitch (nm) Pitch (nm)

EE3230 Ping-Hsuan Hsieh 40


Wire Engineering
• Goal: to achieve delay, area, and power goals with
acceptable noise
• Degrees of freedom:
– Width 2.0 0.8
1.8 0.7
– Spacing

Coupling:2Cadj / (2C adj+Cgnd)


1.6
0.6
1.4 WireSpacing
(nm)
Layer
0.5

Delay (ns):RC/2

1.2
320
1.0 0.4 480
640

Shielding
0.8
– 0.6
0.3

0.2
0.4
0.1
0.2
0 0
0 500 1000 1500 2000 0 500 1000 1500 2000
Pitch (nm) Pitch (nm)

vdd a0 a1 gnd a2 a3 vdd vdd a0 gnd a1 vdd a2 gnd a0 b0 a1 b1 a2 b2

EE3230 Ping-Hsuan Hsieh 41


Repeaters
• R and C are proportional to l (length)
• RC delay is proportional to l2
– Unacceptably long delays for long wires
• Break long wires into N shorter segments
– Drive each one with an inverter or buffer
Wire Length: l

Driver Receiver

N Segments
Segment
l/N l/N l/N

Driver Repeater Repeater Repeater Receiver

EE3230 Ping-Hsuan Hsieh 42


Repeater Design
• How many repeaters should we use?
• How large should each one be?
• Equivalent Circuit
– Wire length l
• Wire Capacitance Cw*l, Resistance Rw*l
– Inverter width W (nMOS = W, pMOS = 2W)
• Gate Capacitance C’*W, Resistance R/W

EE3230 Ping-Hsuan Hsieh 43


Repeater Design
• How many repeaters should we use?
• How large should each one be?
• Equivalent Circuit
– Wire length l/N
• Wire Capacitance Cw*l/N, Resistance Rw*l/N
– Inverter width W (nMOS = W, pMOS = 2W)
• Gate Capacitance C’*W, Resistance R/W

EE3230 Ping-Hsuan Hsieh 44

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy