3d-integrated circuits: a focus on signal ...terrazon process with dbi & flip chips process...
TRANSCRIPT
-
1
1
3D-INTEGRATED CIRCUITS: A FOCUS ON SIGNAL INTEGRITY AND
ELECTROMAGNETIC COMPATIBILITY
Etienne SICARDINSA/DGEIUniversity of Toulouse31077 Toulouse - [email protected]
www.ic-emc.org
2
Cassoulet (heavy food)
Rugby (heavy efforts)
Airbus A380 (heavy airplane)
TOULOUSE - FRANCE
Founded in -120 B.C (heavy history)
Best place to study in France (2011 ranking)(heavy responsibility)
-
2
3
I. EMC ISSUES
II. EVOLUTION OF ICS AND CONSEQUENCE ON EMC
III. 3D IC TECHNOLOGY
IV. MEASUREMENT OF 3D-IC EMISSION AND
SUSCEPTIBILITY
V. MODELS FOR 3D-EMC SIMULATION
VI. DESIGN GUIDELINES FOR IMPROVED EMC
SUMMARY
4
I. EMC ISSUES
-
3
5
5 October 12
http://www.interferencetechnology.com > markets > news
GENERAL EMC ISSUES – ALL DOMAINS
6
Mobile phonePersonal entrainments
Safety systemsComponents
SUSCEPTIBILITY
EquipementsCarbon airplane
Boards
Radar
INDUSTRIAL PRESSURE
Control Systems
EMISSION
-
4
77 October 12
EMC AT IC LEVEL
WHEN EMC OF ICS STARTED?
• Until mid 90’s, IC designers had no consideration about EMC problems in their design..
• Starting 1996, automotive customers started to select ICs on EMC criteria
• Starting 2005, mobile industry required EMC in System in package
• Starting 2015, massive 3D integration will require careful EMC design
• “Urgent Need to Integrate EMC and Product Safety into Engineering Curriculum of Technical Universities”
8
10k
100k
1M
1996 2000 20021998 2004
GSM 14k
GPRS 64k
Year
2006
Bandwidth (bit/s)
2008
10M
100M
Average bandwidth
HSDPA 1.0 M
HSUPA 4 M
2010
LTE 20 M
UMTS 120 k
We are Here
Boundaries
2G(4.5 Billion users)
3G(1 Billion)
3G+(1 Billion)
4G(0.1 Billion)
3D-IC DRIVING FORCES
THE GIANT SCALE MOBILE PHONE INDUSTRY
2012
LTE-A 40 M
2014
-
5
9
3D-IC DRIVING FORCES
4G REQUIRES 4 CAMERAS, 10+ PROCESSORS, GB MEMORY• Texas Instruments OMAP 5430 example• 3D Ics are used for memories and camera modules
• 2008 : “Why 3D?”• 2010 “”How 3D? • 2012 : “When 3D?”• …• 20xx : “Why 2D?”
10
EMC IN 3D-ICS
Georgia-Tech vision of SoC
10 October 12
From Georgia Tech 3D system packaging research http://www.prc.gatech.edu.
HOW TO ENSURE EM COMPATIBILITY OF THIS ..?
-
6
11
Emission and immunity levels (dB)
-40
-30
-20
-10
0
10
20
30
40
50
1 10 100 1000
Frequency (MHz)
Interference risk
Security margin Small margin
Immunity
Immunity
Emission
Emission
EMC MARGIN
MIXING EMISSION AND IMMUNITY LEVELS
FREESCALE « 3G phone in a package » using RCP technology
12
Why a margin
Domain Life time Margin
Aeronautics 30 years 40 dB
Automotive 15 years 20 dB
Mobile phone 1 year 0 dB
Margin depends on application
Acceptable noise level
Process variations
Measurement error
Ageing
Required noise level
Environment
Security
EMC MARGIN
LINK BETWEEN MARGIN AND APPLICATIONS
Ioff/Ion MOS 32-nm
PhD A. C. Ndoye, INSA, 2010
Immunity vs. ageing (LTOL)
-
7
13
dBµV
0
20
40
60
80
100
10 100 1000
FM GSMRF
EMC IS CRITERION FOR SELECTING IC SUPPLIERS
Supplier A
EMC compliant
Not EMC compliant
Customer's specified limit
LOW EMISSION = DIFFERENTIATOR
B
A
Supplier B
14
Thunderstorm impact
UNINTENTIONAL ELECTROMAGNETIC SOURCES
TV UHF
Radars
2-4G BS
1W
Frequency
1MW
1KW
1GW Weather Radar
3 MHz 30 MHz 300 MHz 3 GHz 30 GHz 300 GHz
Power
1mW
HF VHF UHF SHF xHF THF
3G
TV VHF
2G4G
• Fields radiated by electronic devices
• Continuous waves & pulsed waves
25m 25mm0.25m 2.5mm2.5m λλλλλλλλ/4 (ideal antenna)0.25mm
SUSCEPTIBILITY ISSUES
-
8
15
II. EVOLUTION OF IC TECHNOLOGY AND CONSEQUENCE ON EMC
16
Technology
Complexity
Packaging
2004
130nm
100M
Core+ DSP1 Mb Mem
Embedded
blocks
2006
90nm
250M
Core DSPs10 Mb Mem
2008
45nm
500M
Dual coreDual DSP RFGraphic Process.100 Mb MemSensors
2010
32nm
2G
Quad CoreQuad DSP 3D Image ProcCrypto processorReconf FPGA, Multi RF1 Gb Memories Multi-sensors
22nm
2012
7G
10 GIGA-DEVICE ICS
5nm
150 G
2020
?
HIGHER COMPLEXITY
-
9
17
HIGHER PARASITIC EMISSION
INCREASED SWITCHING DEVICE PERFORMANCES
Current drive
(mA/µm)
2.0
1.0
0.0
130 nm
1.5
0.5
45 nm65 nm 22 nm32 nm 17 nm
Technology node
Intrinsic performances
Strain
Gate material
90 nm
Strain to increase mobility
High K Metal Gate to increase field
effect
Tri-Gate for increasing drive
current and reducing leakage
18
Time
WHY TECHNOLOGY PROGRESSES INCREASE EMISSION
Time
• The current amplitude is a little reduced
• The switching is faster
The noise is increased
The noise is increased
New process
Volt
Old process
CurrentOld process
di/dt
New process
Vss
Vdd
t
iLV
∆∆=∆
t
iLV
∆∆=∆
HIGHER PARASITIC EMISSION
-
10
1919 October 12
INCREASED SUSCEPTIBILITY
100 mV margin
5.0
3.3
2.5
1.8
0.5µ 0.35µ 0.18µ 90n 65n
Technology
1.0
Supply (V)
1.2
45n
I/O supply
Core supply
32n
DECREASED NOISE MARGIN IN ICS
22n 17n130n
500 mV margin
Adapted from ITRS roadmap for semiconductors, 2011
20
3D-IC EMC CHALLENGES
20 October 12
J. Kim; IEEE EMC Society Distinguished Lecturer Seminar: Signal Integrity of TSV-Based 3D IC
BASIC EMC ISSUES IN 3D-ICS
-
11
21
III. 3D-IC TECHNOLOGY
22
3D-IC DRIVING FORCES
3D-IC DRIVING FORCES
MEMORIES AND IMAGERS
From 3DIC & TSV Report Cost, Technologies & Markets, 2007, Yole Dev.
-
12
23
3D-IC BENEFITS
3D TECHNOLOGY
Enables the integration of ics fabricated in different technologies : Cmos, CCD, SOI, sensors, MEMS
23 October 12
C. Bower, et. al., “High Density Vertical Interconnects for 3D Integration of Silicon ICs,” 56th ECTC, San Diego, 2006.
B. Aull, et. al., “Laser Radar Imager Based on 3D Integration of Geiger-Mode Avalanche Photodiodes” IEEE SSCC 2006.
4µm viasBosch process
0.18µm SOI0.35 µm SOISensor
24
3D-IC TECHNOLOGY
24 October 12
TERRAZON PROCESS WITH DBI & FLIP CHIPS
Process accessible by CMC, CMP and MOSIS for academics, SME and industries
From CMP annual users meeting, “3D-IC Integration”, January 20th 2011, PARIS
Direct bond interconnect
-
13
25
3D-IC EXAMPLE
25 October 12
Development of 3D Integrated Circuits for HEP, R. Yarema, 2006
3D IC Pixel Electronics, the Next Challenge, R. Yarema, 2008
26
3D-IC EFFICIENCY
IMPROVED ELECTRONIC EFFICIENCY
• 3D reduces interconnect length > inductance effects
• 3D simplifies multiple supply voltage distribution
• 3D reduces package pin count
• More uniform, high density power delivery
26 October 12
J. Lu, “Monolithic 3D Power Delivery Using Dc-Dc Converter”, 3D Architecture Conference, October, 2006, Burlingame, CA.
-
14
27
3D-IC BENEFITS
DIE BONDING VS THROUGH SILICON VIA
27 October 12
EMC-3D Consortium Overview and CoO Model, Paul Siblerud, www.emc3d.org
Ω≈≈ 500 Cl
ZΩ≈≈ 2000 cL
Z
28
SIGNAL INTEGRITY IN 3D-ICS
28 October 12
A SIGNIFICANT DECREASE OF OVER/UNDERSHOOTS
2D 3D
Buffer 3-stage 1-stage
Pad Load 3-5 pF 1 pF
Interconnectcapa
5-20 pF 0.1-5 pF
Interconnectinductance
5-30 nH 0.1-2 nH
Current drive 10-100 mA 1-10 mA
HDI PCB
Stacked die bonding
3D TSV
-
15
29
SIGNAL INTEGRITY IN 3D-ICS
A SIGNIFICANT REDUCTION IN I/O COMPLEXITY
• Most Electrostatic Discharge, overstressprotections can be removed
29 October 12
Solder ballESD protection
Voltage translation and level shifers
30
POWER EFFICIENCY 3D-ICS
A BETTER POWER EFFICIENCY
• Shorter wires decrease the average parasitic inductance and resistance and decrease the number of repeaters needed for long wires.
• Capacitance is reduced if the dies are thinned (350 > 10-100 µm)
• The wire efficiency is improved by 10-100 %
• Power dissipation may be reduced by a factor of 2-10
30 October 12
T. Topol « Three-dimensional integrated circuits », Ibm Journal Research, 2006
Lili Zhou, "Implementing a 2-Gbs 1024-bit ½-rate Low-Density Parity-Check Code Decoder in 3D-Ics « , ICCD 2007
-
16
31
3D-IC INTERFACING
NO I/O STANDARDIZATION BETWEEN INTERFACES
• Cost-effective high-volume manufacturing will be difficult to achieve unless manufacturing standards are developed
• Identify and create new standards in 3D-ICs.
� Through silicon via size
� Interposer and die thickness
� Microbump dimensions
� Electrical behavior of I/Os
• No I/O standardization at present
• IBIS could play an important role for standard interfacing
31 October 12
From 3DIC & TSV Report Cost, Technologies & Markets, 2007, Yole Dev.
From ITRS RoadmapFrom http://www.semi.org/node/37306
32
IV. MEASUREMENT OF 3D-IC EMISSION & SUSCEPTIBILITY
-
17
33
3D-IC MEASUREMENT METHODS
33 October 12
IEC – International Electro-Technical Commission - www.iec.ch
International ElectrotechnicalCommission
169 technical committee
TC 4 : hydraulic turbinesTC 4 : hydraulic turbines
TC 20 : electric cablesTC 20 : electric cables
SC 46A : coaxial cables
SC 47A : integrated circuits
SC 47A : integrated circuits
.
.
.
WG 9 : Test procedures and measurement methods for EMC of ICs
Members
USA
Japan
South Korea
Poland
England
Germany
France
Belgium
Italy
Netherlands
SC 47A : integrated circuits
Secretary : Japan
Work Group
WG 2 : Digital integrated circuits
WG 4 : Interface integrated circuits
WG 7 : Advanced hybrid integrated circuits
WG 9 : Test procedures and measurement methods for EMC of ICs
WG 2 : Digital integrated circuits
Members
USA
Japan
South Korea
Poland
England
Germany
France
Belgium
Italy
Netherlands
Worldwide expert meetings, defending national industrial approaches
3434 October 12
3D-IC MEASUREMENT METHODS
STANDARD EMC TEST BOARD – 2D/3D DOES NOT MAKE DIFFERENCE
100 mm
MetallizationVia row
DUT
Additional holes
Top layer – DUT layer
Bottom layer – Additional component
Connections of IC by 0.25mm vias
Ground plane under IC Connection
layer 1 to 4 by 0.8mm vias
Decoupling of supply on this part of ground plane
Tinned
Layer 1 – Ground
Layer 2 – Power supply
Layer 3 – Signal
Layer 4 – Ground and/or signal
-
18
35
STANDARD IC EMISSION MEASUREMENT
METHODS - IEC 61 967
IEC 61967-2(TEM : 1 GHz)
IEC 61967-3/6(Near-field Scan, 5 GHz)
IEC 61967-4(1/150 Ω, 1 GHz)
IEC 61967-8(Mini-stripline)
IEC 61967-7(Mode Stirred Chamber : 18 GHz)
Ext : IEC 61967-2(GTEM 18 GHz)
3D-IC MEASUREMENT METHODS
36
3D-IC Measurement Methods
36 October 12
IEC 61 967 – 2 “TEM/GTEM CELL” -8 “IC-STRIPLINE”
� Upper IC may play the role of shielding
� Upper IC is closer to septum
� 45 mm for TEM, 6.7 mm to Stripline
Low SSN on top High SSN on top
TEM/GTEM cross-section
IC-Stripline cross-section
TEM/GTEM cross-section
45 mm
6,7 mm
-
19
37
3D-IC MEASUREMENT METHODS
37 October 12
IEC 61 967 – 3/6 “NEAR-FIELD SCAN”
� Combining laser and high-precisionNFS may lead to precise 3D investigations
� But lower die stack shielded by upperdies
� Faster 3D scan: may use « cube-probe » (isotropic measurement of Hx, Hz, Hz) for Hmax measurement
D. Baudry, PhD report, ESIGELEC, Univ Rouen, 2005
38
3D-IC MEASUREMENT METHODS
38 October 12
IEC 61 967 – 4 “1/150 Ω METHOD”
Sub-component
Passive Distribution Network
IB
Immunity behavioral
PDN
Σ,Π
Silicon Die
IA
Internal activity
PDN Embedded passives
PDN package
Other silicon die
Other sub-component
Integrated Circuit
� Each die would have a built-in 1 Ωprobing for IC emissioncharacterization
1 Ω die 11 Ω die 2
1 Ω die 3
-
20
3939 October 12
IEC 62132-3(Bulk Current Injection : 1 GHz)
IEC 62132-4(Direct Power Inj 1GHz)
IEC 62132-2(TEM/GTEM)
IEC 62132-5(WBFC 1 GHz)
IEC 62132-6(LIHA : 10 GHz)
IEC 62132-8 Mini Stripline
INTERNATIONAL STANDARDS FOR IC SUSCEPTIBILITY MEASUREMENTS
3D-IC MEASUREMENT METHODS
40
3D-IC MEASUREMENT METHODS
40 October 12
3D-RELEVANT IMMUNITY MEASUREMENT
METHODS (RESEARCH)
Skate-probe On-chip sampling
Analog JTAG DPI+NFS
-
21
41
3D-IC MEASUREMENT METHODS
41 October 12
IEC 62 132 – 3 “BCI METHOD”� BCI is 5 Kg, 20 cm
diameter
� 3D-IC is 3g, 25 mm square
3D-IC connected to a bus (CAN, LIN..) or sensors ?
42
3D-IC MEASUREMENT METHODS
42 October 12
IEC 62 132 – 4 “DPI METHOD”
Sub-component
Passive Distribution Network
IB
Immunity behavioral
PDN
Σ,Π
Silicon Die
IA
Internal activity
PDN Embedded passives
PDN package
Other silicon die
Other sub-component
Integrated Circuit
� Each die would have built-in injection probes for IC immunity characterization
to die 1to die 2
to die 3
http://ecubes.epfl.ch
-
22
43
3D-IC MEASUREMENT METHODS
43 October 12
IEC 62 132 – 4 “DPI METHOD”
� Extension of injection to 10 GHz « x-DPI » concept
PCB Board level
1 mm
Interposer level
100 µm
DC
RFITSV to IC
DPI Capa (10-100 pF)
DC Polarization and Isolation resistance(or embedded L)
44
3D-IC MEASUREMENT METHODS
44 October 12
IEC 62 132 – 2 “TEM/GTEM METHOD”, - 8 “MINI-STRIP LINE”
� Mini strip-line efficient for injection (3%)
� Canonical field
F. Klotz, « IC-Stripline, new method for emission and immunity », EMC Compo 2009
EMC test board
Strip line cross-section
6.7 mm
-
23
45
3D-IC MEASUREMENT METHODS
45 October 12
SKATE PROBE PRINCIPLES� Investigate “what-if
3D” by manual positioning in X,Y,Z
� Avoid the cost of 3D prototypes
� Ensure EMC prior to 3D-IC design
Characterization of the aggressor NF emission
The aggressor SkateProbedesign
SkateProbe validation
Measurement of the coupling between the SkateProbe and the victim
Victim component
An Innovative Methodology for Evaluating Multi-Chip EMC in Advanced 3G Mobile Platforms, S. Akue Boulingui, IEEE EMC Symp Austin, 2009
RF Stage Numerical part
Power management
MemoryProcessor3G Transceiver3G PA
3G LNA
2G PAs
2G Transceiver
46
3D-IC MEASUREMENT METHODS
46 October 12
SKATE PROBE – CASE STUDY� Aggressor : 3G Power amplifier
� Victim : 3G transceiver
-
24
47
3D-IC MEASUREMENT METHODS
47 October 12
PA_SkateProbeSkateProbe signal Transceiver
Power amplifier
Desired signalUSB conexion
Power meter
-6 dB
Coupler
SKATE PROBE – CASE STUDY� 3D positioning of the
aggressor above the victim� SNR characterization
with/without disturbances
-8,5
-7,5
-6,5
-5,5
-4,5
2110 2120 2130 2140 2150 2160 2170
SN
R (
dB)
SNR targetWith
disturbance
SNR referencecurve
Frequency (MHz)
SN
R (
dB)
50 Ω load
11 mm
14 mm
48
3D-IC MEASUREMENT METHODS
48 October 12
ON-CHIP SAMPLING � Simple Sample/Hold analog cell
to probe voltages & currents
Sampling command
S/H cell
Signal to measure
+
_
Output amplifier
Sampled dataHigh impedance
probe
Attenuator
Sensor
S. Ben Dhia, “On-Chip Noise Sensor for Integrated Circuit Susceptibility Investigations”, IEEE Trans. Instr. Meas, 2012
Enrique LAMOUREUX, PhD2006, INSA Toulouse, France
-
25
49
3D-IC MEASUREMENT METHODS
49 October 12
ON-CHIP SAMPLING
� Setup with DPI
� Clock and Data trig the IC under test
Switching
High frequency Low frequency
Switching Switching
50
3D-IC MEASUREMENT METHODS
50 October 12
-50
-40
-30
-20
-10
0
10
20
30
40
1,E+06 1,E+07 1,E+08 1,E+09
Frequency (Hz)
EM
I tra
nsfe
r fun
ctio
n (d
B V
/W) Simulation
Measurement
ON-CHIP SAMPLING� Extraction of the
transfer function of the IC, strongly dependant on F
� On-chip sensing bandwidth linked to technology
� 2 GHz 0.25µm, 20 GHz 22 nm
S. Ben Dhia, “On-Chip Noise Sensor for Integrated Circuit Susceptibility Investigations”, IEEE Trans. Instr. Meas, 2012
-
26
51
3D IC Measurement Methods
51 October 12
ANALOG JTAG� Usable with digital JTAG
boundary scan� Adds analog testability –
both controllability and observability
� Eliminates large area needed for analog test points
Agrawal, “IEEE 1149.4 JTAGAnalog Test Access Port and Standard”, VLSI Test Lecture 2001
52
3D IC MEASUREMENT METHODS
52 October 12
COMBINED DPI AND NFS� Idea: inject in direct
power injection and measure in Near-field scan
� Tool for observing the 3D-IC from outside
� May help investigating the coupling paths and build accurate models
A. Alaeldine “Analysis of the Propagation of EM Disturbances Inside Integrated Circuits Using DPI and NFS”, IEEE EMC Austin 2009
DPI off
DPI on
-
27
53
V. MODELS FOR EMC SIMULATION
54
DESIGN
Architectural Design
Design EntryDesign Architect
FABRICATION
EMC compliant
EMC SimulationsCompliance ?
GO
NO
EMC VALIDATED BEFORE FABRICATION
Design Guidelines
Tools
Training
EMC-AWARE DESIGN CYCLE
-
28
55
IEC STANDARD MODEL APPROACH
1. ICEM-CE - Conducted RF emission
2. ICEM-RE - Radiated RF emission
4. ICIM-CI - Conducted RF immunity
4. ICIM-RI - Radiated RF immunity
www.iec.ch
IEC 62 433 – EMISSION (ICEM) AND IMMUNITY (ICIM) MODELS
Conducted mode models in industrial use
56
GENERIC FLOW
Core – I/O ModelPackage ModelTest board ModelTest bench Model
EMC Model for the circuit
Electrical Simulation
Simulated Emission spectrum
EMC SIMULATION FLOW AT IC LEVEL
The DUT is isolated on a simple EMC board to minimize modeling effort
-
29
57
Core Model
Probe Model Analog Time-Domain
Simulation
Fourier Transform
Conversion to Win-SPICE
dB vs Freq (log)conversionPackage
Model
Test board Model
INFINEON TRICORE™ - TEM CELL MODEL
EMISSION CASE STUDY
Capacitance coupling to the TEM cell
0A
0.5A
1.0A6.6ns (150MHz))
Core current model
58
Infineon TriCore™ measurement/simulation comparisons
Measure� Correct envelop
� Reasonable match
� Simulation 15 dB above measurement starting 700 MHz
� Manual fit leads to 5 dB max difference
Radiated noise in GTEM cell (dBµV)
INFINEON TRICORE™ - TEM CELL EMISSION
EMISSION CASE STUDY
Predictive model
-
30
5959
Sub-component
Passive Distribution Network
IB
Immunity behavioral
PDN
Σ,Π
Silicon Die/ Intellectual property
IA
Internal activity
PDN Embedded passives
PDN package
Other silicon die
Other sub-component
Integrated Circuit (ICEM, ICIM)
Integrated Circuit (ICEM, ICIM)
Sub-component
Silicon die
DOWN-SCALING TO MULTI-DIES
EMISSION MODEL OF MULTI-DIE 2D/3D ICS
One-die to multi-die model
60
3D-IC MODEL
THROUGH SILICON VIA MODEL
60 October 12
J. Kim; IEEE EMC Society Distinguished Lecturer Seminar: Signal Integrity of TSV-Based 3D IC
• TSV diameter : 55 µm• TSV SiO2 thickness : 0.5 µm
Low loss upto 30 GHz
-
31
61
3D-IC MODEL
61 October 12
Resistance and Capacitance values in mΩ and fF
M. B. Healy, “A Study of Stacking Limit and Scaling in 3D ICs: An Interconnect Perspective”, 2009 Electronic Components and Technology Conference
THROUGH SILICON VIA MODEL
Large capacitance for large & long TSVs
6262
M. B. Healy, “A Study of Stacking Limit and Scaling in 3D ICs: An Interconnect Perspective”, 2009 Electronic Components and Technology Conference
3D-IC MODEL
THROUGH SILICON VIA MODEL
PAK et al.: Pdn impedance modeling and analysis of 3D TSV IC, IEEE Trans on components, packaging, and manuf. tech, Feb 2011
• Tall 3D Ics lead to significant L. • Crosstalk should also be considered
-
32
63
3D-IC DELAY MODELING
DELAY VS. INTER-DIE OPTIONS
63 October 12
F2F F2B
J. Roulard, Electrical Characterization and Impact on Signal Integrity of New Basic Interconnection Elements inside 3D Integrated Circuits, 2011 Electronic Components and Technology Conference
• Stacking strategy may reduced delay by 80%
64
3D-IC EMC PERFORMANCES
EMISSION PREDUCTION
64 October 12
• Long bonding act as antennas
• l/4 eq. 5-15 GHz • Important conducted,
radiated noise
µC
mem
Package
• Mem acts as a shielding• Very fast µC/mem
exchanges• Still important antenna• Medium emission
(conducted/radiated)
• Shielding of mem, µC• No antenna effect• TSV may act as a load and
slow down mem/µc exchanges
• Low emission
-
33
65
3D-IC MODEL
EMISSION PREDICION
E. Sicard, Wu Jianfei www.ic-emc.org “3D-IC Case study”
• 3D-TSV technology may reduce emission by 30 dB vs PCB, and 15 dB vs die stack with wire bonding
-30 dB
Frequency
Radiated emission (dBµV)
66
3D-IC POWER INTEGRITY
POWER INTEGRITY MODELING
66 October 12
Young-Joon Lee, Co-design of Reliable Signal and Power Interconnects in 3D Stacked Ics, 2009
Maximum power noise (mV) with varied settings
Power delivery by specific TSVs
-
34
67
3D-IC POWER INTEGRITY
67 October 12
Voltage drop as a function of time. The dynamic noise as a function of TSV dimension in µm (40 x 40 TSV)
M. B. Healy, “A Study of Stacking Limit and Scaling in 3D ICs: An Interconnect Perspective”, 2009 Electronic Components and Technology Conference
POWER INTEGRITY MODELINGCore placement strategy has a direct impact on voltage drops
68
Package
IB
PDN
Package PDN
Silicon die
IC PDN Internal BehaviourIB
External pins
ICIM – immunity modelPackage
Monitoring of the failure
PDN = Passive Distribution Network
detection
RF disturbance Coupling path
Close to ICEM-CE
Close to ICEMAdd Diodes (camp, back-to-back, ESD, EOS)
New!
IEC 62433-4 – “ICIM CONDUCED IMMUNITY”
IEC STANDARD MODEL APPROACH
Based on ICEM, add non-linearities
-
35
69
• 16 bit micro-controller• Direct power injection• Input buffer aggression• Sinusoidal mode• Simulation criterion:
Logical change of input buffer
From A. Boyer’s PhD, INSA, 2007
IMMUNITY SIMULATION
S12X CASE STUDY – DPI ON AN INPUT
70
VI. DESIGN RULES FOR IMPROVED EMC
-
36
71
Why: because inductance is a major source of resonanceWhere is the inductance: in each conductor, worst is far from ground
L=1 nH/mm
DESIGN RULES
GUIDELINE 1 : REDUCE THE INDUCTANCE
ICIC
VDD
VSS
VSS
VDD
72
Multiple VDD, VSS Tools required to forecast strong di/dt effects
DESIGN RULES
GUIDELINE 2 : PLACE VDD/VSS CLOSE TO STRONG DI/DT
-
37
73
• to reduce current loops that provoke magnetic field• to increase decoupling capacitance that reduces fluctuations
DieLeadLead
current
Added contributions
Canceled contributions
EM wave
GUIDELINE 3 : PLACE VDD-VSS CLOSE
DESIGN RULES
Added
Canceled (-20 dB)
74
Correct Fail
9 I/O ports
DESIGN RULES
GUIDELINE 4 : USE ONE VDD/VSS FOR 10 I/OS
• To reduce current loops• To reduced LC effects
BGA examples
3D-TSV examples
J. S. Pak “PDN Impedance Modeling and Analysis of 3D TSV IC », IEEE Transactions on components, packaging, and manuf. Tech. vol. 1, no. 2, Feb. 2011
Both correct
-
38
75
Poor design:x 5 mode switching noise
© Dr. Howard Johnson, "BGA Crosstalk", www.sigcon.com
DESIGN RULES
GUIDELINE 4 : USE ONE VDD/VSS FOR 10 I/OS FPGA CASE STUDY
7610/22/2012 Graduate Student Meeting on Electronic
Engineering - Tarragona76
• 1nF added close to the core• More than 15 dB noise reduction• Selected areas:
• Logic Core• Charge pump (FLASH)• Fast I/Os (DDRx)
B.Vrignon CESAME test-chip IEEE Trans EMC 2006
GUIDELINE 5 : PLACE ON-CHIP DECOUPLING CLOSE TO STRONG DI/DT
DESIGN RULES
-
39
77
Filler-cap at the output buffer area
High speed port
High speed port
Protection circuit
VDD
VSS
Core
Pad
Protection circuit
Filler-cap
GUIDELINE 5 : PLACE ON-CHIP DECOUPLING CLOSE TO STRONG DI/DT
DESIGN RULES
78
GUIDELINE 6 : ADD JIITTER ON THE CLOCK
© B. Vrignon, Freescale SAS
-8 dB
DESIGN RULES
-
40
79
RC filtering works both for emission and immunity
Ali ALAELDINE , PhD Eseo France
Susceptibility level
Normal Core
Isolated Core
RC Code
GUIDELINE 7 : ADD RC FILTERING TO ISOLATE NOISE
DESIGN RULES
Frequency (MHz)
80
� « Zero » emission designs
� cancel all fields
� negligeable external di/dt
� De-synchronized parts
� « Zero » susceptibility designs
� Clamp, filter
� Protect, shield by design, materials
� Defensive core
supply
Critical sensor
Critical decision
� Reduce frequency� Reduced VDD� Smart capa management� Redundancy
RFI, ESD, EOS, EFT
Detector
… and Why Not?INNOVATIVE EMC …
FUTURE
-
41
81
CONCLUSION
82
• Higher complexity and frequencies, technology scale down make EMC more and more challenging
• EMC is still investigated late in the design flow
• Mature standard measurement methods dedicated to ICs
• New standards for EMC modeling at IC level
• Good prediction of emission and susceptibility up to 2 GHz for 2D-ICs• 3D-ICs speed up signal propagation, consumes less power• Many 3D-IC technologies co-exist, no standard• New EMC Challenges in 3D due to die-die proximity, protection
simplification• Signal Integrity closely linked to via technologies • PDN & Power Integrity linked to 3D choices• Standard and new measurement methods available• EMC impacted by positioning, designs, filtering and assembly options• 3D-EMC still in infancy stage, a huge room for innovation
CONCLUSION
-
42
83
Standards www.iec.ch
• IEC 61967, 2001, Integrated Circuits emissions
• IEC 62132, 2003, integrated circuits immunity
• IEC 62433, 2006, Integrated Circuit Model
• IEC 62215, 2009Transcient immunity
Books
www.springeronline.com
REFERENCES
www.emccompo.org
Workshops
EMC Compo in Dec 13Nara
Japan
Tools
www.ic-emc.org
Melbourne, Australia from 20 – 23 May 2013www.apemc13.org
84
Merci beaucoup pour votre attentionThank you very much for listing
• The audience• The IEEE Solid State Circuits Society, Switzerland• The IEEE Electron Device Society, Switzerland• The IEEE Student Branch, EPFL• Wladyslaw Grabinski• Lucian Barbut