digital integrated circuits lecture 20: package, power, clock, and...
TRANSCRIPT
DIC-Lec20 [email protected] 1
Digital Integrated CircuitsLecture 20: Package, Power, Clock, and I/O
Chih-Wei Liu
VLSI Signal Processing LAB
National Chiao Tung University
DIC-Lec20 [email protected] 3
Packages
Package functionsElectrical connection of signals and power from chip to boardLittle delay or distortionMechanical connection of chip to boardRemoves heat produced on chipProtects chip from mechanical damageCompatible with thermal expansionInexpensive to manufacture and test
DIC-Lec20 [email protected] 5
Multichip Modules
Pentium Pro MCMFast connection of CPU to cacheExpensive, requires known good dice
DIC-Lec20 [email protected] 6
Chip-to-Package Bonding
Traditionally, chip is surrounded by pad frameMetal pads on 100 – 200 μm pitchGold bond wires attach pads to packageLead frame distributes signals in packageMetal heat spreader helps with cooling
DIC-Lec20 [email protected] 7
Advanced Packages
Bond wires contribute parasitic inductanceFancy packages have many signal, power layers
Like tiny printed circuit boardsFlip-chip places connections across surface of die rather than around periphery
Top level metal pads covered with solder ballsChip flips upside downCarefully aligned to package (done blind!)Heated to melt ballsAlso called C4 (Controlled Collapse Chip Connection)
DIC-Lec20 [email protected] 8
Package Parasitics
Chip
Signal P
ins
PackageCapacitor
Signal P
ads
ChipVDD
ChipGND
BoardVDD
BoardGND
Bond Wire Lead Frame
Package
Use many VDD, GND in parallelInductance, IDD
DIC-Lec20 [email protected] 9
Heat Dissipation
60 W light bulb has surface area of 120 cm2
Itanium 2 die dissipates 130 W over 4 cm2
Chips have enormous power densities
Cooling is a serious challenge
Package spreads heat to larger surface areaHeat sinks may increase surface area further
Fans increase airflow rate over surface area
Liquid cooling used in extreme cases ($$$)
DIC-Lec20 [email protected] 10
Thermal Resistance
ΔT = θjaPΔT: temperature rise on chip
θja: thermal resistance of chip junction to ambientP: power dissipation on chip
Thermal resistances combine like resistorsSeries and parallel
θja = θjp + θpa
Series combination
DIC-Lec20 [email protected] 11
Example
Your chip has a heat sink with a thermal resistance to the package of 4.0° C/W. The resistance from chip to package is 1° C/W.The system box ambient temperature may reach 55° C.The chip temperature must not exceed 100° C.What is the maximum chip power dissipation?
(100-55 C) / (4 + 1 C/W) = 9 W
DIC-Lec20 [email protected] 12
Power Distribution
Power Distribution Network functionsCarry current from pads to transistors on chip
Maintain stable voltage with low noise
Provide average and peak power demands
Provide current return paths for signals
Avoid electromigration & self-heating wearout
Consume little chip area and wire
Easy to lay out
DIC-Lec20 [email protected] 13
Power Requirements
VDD = VDDnominal – Vdroop
Want Vdroop < +/- 10% of VDD
Sources of Vdroop
IR dropsL di/dt noise
IDD changes on many time scalesclock gating
Time
Average
Max
Min
Power
DIC-Lec20 [email protected] 14
Power System Model
Power comes from regulator on system boardBoard and package add parasitic R and LBypass capacitors help stabilize supply voltageBut capacitors also have parasitic R and L
Simulate system for time and frequency responses
VoltageRegulator
Printed CircuitBoard Planes
Packageand Pins
SolderBumps
BulkCapacitor
CeramicCapacitor
PackageCapacitor
On-ChipCapacitor
On-ChipCurrent DemandVDD
Chip
PackageBoard
DIC-Lec20 [email protected] 15
Bypass Capacitors
Need low supply impedance at all frequenciesIdeal capacitors have impedance decreasing with ωReal capacitors have parasitic R and L
Leads to resonant frequency of capacitor
104
105
106
107
108
109
1010
10-2
10-1
100
101
102
frequency (Hz)
impedance
1 μF
0.03 Ω
0.25 nH
DIC-Lec20 [email protected] 16
Frequency Response
Use multiple capacitors in parallelLarge capacitor near regulator has low impedance at low frequenciesBut also has a low self-resonant frequencySmall capacitors near chip and on chip have low impedance at high frequencies
Choose caps to get low impedance at all frequencies
frequency (Hz)
impedance
DIC-Lec20 [email protected] 17
Clock Distribution
On a small chip, the clock distribution network is just a wire
And possibly an inverter for clkbOn practical chips, the RC delay of the wire resistance and gate load is very long
Variations in this delay cause clock to get to different elements at different timesThis is called clock skew
Most chips use repeaters to buffer the clock and equalize the delay
Reduces but doesn’t eliminate skew
DIC-Lec20 [email protected] 18
Example
Skew comes from differences in gate and wire delayWith right buffer sizing, clk1 and clk2 could ideally arrive at the same time.But power supply noise changes buffer delaysclk2 and clk3 will always see RC skew
3 mm
1.3 pF
3.1 mmgclk
clk1
0.5 mm
clk2clk3
0.4 pF 0.4 pF
DIC-Lec20 [email protected] 19
Review: Skew Impact
F1 F2
clk
clk clk
Combinational Logic
Tc
Q1 D2
Q1
D2
tskew
CL
Q1
D2
F1
clk
Q1
F2
clk
D2
clk
tskew
tsetup
tpcq
tpdq
tcd
thold
tccq
( )setup skew
sequencing overhead
hold skew
pd c pcq
cd ccq
t T t t t
t t t t
≤ − + +
≥ − +
144424443
Ideally full cycle isavailable for workSkew adds sequencingoverheadIncreases hold time too
DIC-Lec20 [email protected] 20
Cycle Time Trends
Much of CPU performance comes from higher ff is improving faster than simple process shrinksSequencing overhead is bigger part of cycle
0.01
0.1
1
10
100
8038680486PentiumPentium II / III
Spec
Int9
5
1985 1988 1991 1994 1997 2000
1.2 0.8 0.6 0.35 2.0
Process
100
200
500VDD = 5 VDD = 3.3
VDD = 2.5
50Fano
ut-o
f-4 (F
O4)
Inve
rter D
elay
(ps)
0.25
10
100
1000
8038680486PentiumPentium II / III
MH
z
1988 1991 1994 1997 20001985
10
100
1985 1988 1991 1994 1997
8038680486PentiumPentium II / IIIFO
4 in
verte
r del
ays
/ cyc
le50
20
2000
DIC-Lec20 [email protected] 21
Solutions
Reduce clock skewCareful clock distribution network design
Plenty of metal wiring resources
Analyze clock skewOnly budget actual, not worst case skews
Local vs. global skew budgets
Tolerate clock skewChoose circuit structures insensitive to skew
DIC-Lec20 [email protected] 23
Clock Grids
Use grid on two or more levels to carry clockMake wires wide to reduce RC delayEnsures low skew between nearby pointsBut possibly large skew across die
DIC-Lec20 [email protected] 24
Alpha Clock Grids
PLL
gclk grid
Alpha 21064 Alpha 21164 Alpha 21264
gclk grid
Alpha 21064 Alpha 21164 Alpha 21264
DIC-Lec20 [email protected] 25
H-Trees
Fractal structureGets clock arbitrarily close to any pointMatched delay along all paths
Delay variations cause skewA and B might see big skew A B
DIC-Lec20 [email protected] 26
Itanium 2 H-Tree
Four levels of buffering:Primary driverRepeaterSecond-level clock bufferGater
Route aroundobstructions
Primary Buffer
Repeaters
Typical SLCBLocations
DIC-Lec20 [email protected] 27
Hybrid Networks
Use H-tree to distribute clock to many pointsTie these points together with a grid
Ex: IBM Power4, PowerPCH-tree drives 16-64 sector buffersBuffers drive total of 1024 pointsAll points shorted together with grid
DIC-Lec20 [email protected] 28
Input / Output
Input/Output System functionsCommunicate between chip and external world
Drive large capacitance off chip
Operate at compatible voltage levels
Provide adequate bandwidth
Limit slew rates to control di/dt noise
Protect chip against electrostatic discharge
Use small number of pins (low cost)
DIC-Lec20 [email protected] 30
Output Pads
Drive large off-chip loads (2 – 50 pF)With suitable rise/fall timesRequires chain of successively larger buffers
Guard rings to protect against latchupNoise below GND injects charge into substrateLarge nMOS output transistorp+ inner guard ringn+ outer guard ring
In n-well
DIC-Lec20 [email protected] 31
Input Pads
Level conversionHigher or lower off-chip VMay need thick oxide gates
Noise filteringSchmitt triggerHysteresis changes VIH, VIL
Protection against electrostatic discharge
AY
VDDH
VDDLA Y
VDDL
A Y
weak
weak
A
Y
DIC-Lec20 [email protected] 32
ESD Protection
Static electricity builds up on your bodyShock delivered to a chip can fry thin gatesMust dissipate this energy in protection circuits before it reaches the gates
ESD protection circuitsCurrent limiting resistorDiode clamps
ESD testingHuman body modelViews human as charged capacitor
PADR
Diodeclamps
Thingate
oxides
Currentlimitingresistor
DeviceUnderTest
1500 Ω
100 pF
DIC-Lec20 [email protected] 33
Bidirectional Pads
Combine input and output padNeed tristate driver on output
Use enable signal to set directionOptimized tristate avoids huge series transistors
PAD
Din
Dout
En
Dout
En Y
Dout
NAND
NOR
DIC-Lec20 [email protected] 34
Analog Pads
Pass analog voltages directly in or out of chipNo bufferingProtection circuits must not distort voltages
DIC-Lec20 [email protected] 35
MOSIS I/O Pad
1.6 μm two-metal processProtection resistorsProtection diodesGuard ringsField oxide clamps
Out
En
Out
PAD
In
264 Ω 185 Ω
In_bIn_unbuffered
600/3
240
160
90
4020
48