2/19/2016 1 Hierarchical Test for Today’s SOC and IoT Yervant Zorian, 2 Agenda Trends & Challenges Hierarchical Test Solution IP-Level Test Preparation SOC-Level Test Optimization Beyond SOC: IEEE Test Stnds. Life-Cycle Test & Diagnosis New Topics Conclusion
78
Embed
hierarchical test tutorial - DATE-Conference · Source: Cisco Systems, VNI Global Mobile Data Traffic Forecast Update 2013 4 Mobile Source: Business Insider, December 2013. 2/19/2016
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
2/19/2016
1
Hierarchical Test for Today’s SOC and IoT
Yervant Zorian,
2
Agenda
Trends & Challenges
Hierarchical Test Solution
IP-Level TestPreparation
SOC-Level TestOptimization
Beyond SOC:IEEE Test Stnds.
Life-Cycle Test & Diagnosis
New Topics
Conclusion
2/19/2016
2
3
Data, Data, Data Higher Capacity, Faster Transfer, and Lower Cost
Traffic from wireless and mobile devices will exceed traffic from wired devices by 2016.
By 2017, the annual global IP traffic will surpass the zettabyte threshold (1.4 zettabytes).
Source: Cisco Systems, VNI Global Mobile Data Traffic Forecast Update 2013
43,570
55,553
68,892
83,835
101,055
120,643
0
20,000
40,000
60,000
80,000
100,000
120,000
140,000
2012 2013 2014 2015 2016 2017
IP T
raff
ic, P
eta
byt
es
pe
r M
on
th
Global IP Traffic
Source: Cisco Systems, VNI Global Mobile Data Traffic Forecast Update 2013
4
Mobile
Source: Business Insider, December 2013
2/19/2016
3
5
The Emerging Scene!
Infrastructuralcore
Internet of Things
Mobileaccess
Courtesy: J. Rabaey
The Cloud!
6
• Business Apps
• Enterprise Resource Management (ERP) Financials
• TechApps (design, eng, R&D)
• ERP HR
• Collaboration Apps
• Email
• Data analysis/mining apps
• Data Backup/archive
• Help Desk/IT Service Management
• Storage Capacity on-demand
• Application development
• IT Management (server network)
• Mobile Device management
What’s Moving to the Cloud
Cloud Computing Driving Growth of Mega Data
Centers
36% CAGR
2/19/2016
4
7
Cloud 2.0 for the Internet of Things Central Cloud for Ubiquitous Connectivity
Wo
rk C
lou
d
Personal CloudAway or on the move
Central Cloud H
om
e C
lou
d
source: IEEE ISSCC Conference, 2014
WiFi Mobile
8
Reducing Power in Data Centers: System Integration Enables New Micro Servers
Server Power Breakdown
CPU Power~1/3
Other ServerH/W~2/3
CPU architecture
Integration & Innovation
Traditional Server
Micro Server
Traditional Micro Server
64-bit CPUSoC
Micro server SoCs integrate 64-bit CPU, networking, security, storage with targeted workload application
acceleration & high-speed I/Os
64bit CPUSoC
2/19/2016
5
9
(iPhone, ’07)(Galaxy S, ’10)
(Nokia 9000, ’96)
Wireless Phone
• Voice →Voice communication with some applications
IP-based designEnabling system companies to design chips(Apple, Microsoft, Amazon, Google….)
• Assemble componentsfrom parameterized library
– Including
• Integrate
– Configurable processor core
– Memories (RAM, ROM)
– Special-purpose standardblocks (ASSPs)
– Glue logic
• Third-party special-purposelogic / MEMS / MEOS
2/19/2016
9
17
Drivers of Hierarchical Test
• Growing volume of IP in an SOC
• Increasing design complexity w multiple levels of hierarchy
• Exploding digital logic size
• Increasing types of IP blocks to realize standard functions
• Growing use of 3rd party IP
• Increasing use of advanced technologies
• Dispersing design teams globally
• Tight time-to-volume schedules
• Improving test & debug access throughout life cycle
18
• IP test based on direct access increases test time and cost• Ad hoc IP test access not scalable• Reduced pin access, flexible test scheduling & concurrent test needed
1. Growing Use of IP
0
100
200
300
400
500
600
700
800
900
65nm 40nm 28nm 20nm
145
245
455
882
IBS, 2012 July
Number of IP Blocks per Design
2/19/2016
10
19
1. Ad-hoc IP/Core Test Access Can’t Scale
Rely on direct I/O accessRely on direct I/O access
How to Access?
Limited I/O access for larger designs with many IP/cores and hierarchy
Direct I/O access is feasible only for small designs
20
2. Multi-Level Hierarchical Design
- Flat DFT approaches and centralized test management are expensive in area and design time
- Hierarchical DFT access and pattern reuse/ porting required
- Automated hierarchical level test sign-off required
GPU GraphicCore(s)
USB
HD-RAM
PCIe
CPUSub-Chip
Sub-Chip
* Small boxes are Self-Testable Interface IP
ATPGMemory Interface IP, Calibration & Repair
Design Centric Yield Analysis
Silicon Debug & Diagnostics
Memory Self-Test & Repair
Off
-Chi
pO
n-C
hip
AMS IP AMS IP
SRAM
SRAM
SRAM
SRAM Test-time and cost
optimization
CompressionCompression
Interface IPSelf-Test & Calibration
Memory Self-Test & Repair
SRAM
SRAM
Core
SRAMSRAM
Core
SRAMSRAM
2/19/2016
11
21
3. Exploding Digital Logic Size
• Increased test time and power consumption during test• Long test development time for flat designs• Divide & conquer: design partitioning and wrapping needed
4. Heterogeneous IP Test Challenges –Memory/AMS/Legacy IP
UDLScan
UDLScan
MemoryScan Wrapper
Hard IPMux I/O Wrapper
MemoryBIST Wrapper
Mem
oryB
IST + Mux I/O
Wrapper
Hard IPMux I/O Wrapper
Hard IPScan Wrapper
UDLScan
UDLScan
Hard
IPB
IST + Scan Wrapper
26
4. Heterogeneous IP Challenges –Embedded Measurement IP
• Power Management Controllers
• Temperature Sensors
• Radio Tuners
• Measurement Probes
• Clock Generators
2/19/2016
14
27
Functional I/O-only access
makes it difficult to develop test
Temperature Sensors
No standard test Interfaces
Uncontrollable Instruments impacts test
quality
4. Embedded Measurement IP
UDLScan
UDLScan
MemoryScan Wrapper
Hard IPMux I/O Wrapper
MemoryBIST Wrapper
Mem
oryB
IST + Mux I/O
Wrapper
Hard IPMux I/O Wrapper
Hard IPScan Wrapper
UDLScan
UDLScan
Hard
IPB
IST + Scan Wrapper
28
5. Growth in 3rd Party IP Usage to Double
Strong Growth in 3rd Party IP Usage
Shorter Time Window for
New Product Launch
Escalating Design Costs
Increasing Design
Complexity
Source: Gartner, 2013
0% 15% 30% 45% 60% 75% 90%
WiredCommunication
Industrial
Automotive
Data Processing
WirelessCommunication
Consumer
2017
2012
Overall 3rd
party design IP use in 2012
Overall 3rd
party design IP use in 2017
Percentage of 3rd party IP blocks used in SoC design
2/19/2016
15
29
2000 2002 2004 2006 2008 2010 20142012
• Before 32nm, new process was introduced every other year
p‐SiON
HK/MG
* Source : ITRS, Samsung Electronics Co.
Integration
180nm
130nm
90nm
65nm
45nm
28nm32nm
20nm
14nm
Since then, a new process every year
30
IDF – Silicon Leadership
• Manufacturing & Test Quality are critical
• Modeling new types of defects (FinFET)
• Programmable test infrastructure, yield optimization and calibration techniques needed
2/19/2016
16
31
7. Globally Dispersed Design Teams
• Ad-hoc hand-off is error prone and time consuming• Automated sub-chip level integration and hierarchical sign-off
needed
0
50
100
150
200
250
5 816
38
81
165
190
220
IBS, 2012 August
Number of IP Blocks per Design
32
8. Shorter TTM and TTV
• Long development Time for ad-hoc bring up and silicon debug
• Uniform access and use of IEEE test standards reduce bring-up complexity and simplifies silicon debug
UDLScan
UDLScan
MemoryScan Wrapper
Hard IPMux I/O Wrapper
MemoryBIST Wrapper
Mem
oryBIST + M
ux I/O W
rapper
Hard IPMux I/O Wrapper
Hard IPScan Wrapper
UDLScan
UDLScan
Hard
IPBIST + Scan W
rapper
2/19/2016
17
33
0
10
20
30
40
50
60
70
80
90
100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Design Production Ramp-up Volume
YieldLearning Curve
Fab Yield Optimization
Yield Life Cycle CurveYield Life Cycle Curve
Yie
ld A
sses
smen
t (%
)
34
0
10
20
30
40
50
60
70
80
90
100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Design Production Ramp-up Volume
New YieldLearningCurve
YieldLearning Curve
Design Yield Optimization
Fab Yield Optimization
Yield Life Cycle CurveYield Life Cycle Curve
Yie
ld A
sses
smen
t (%
)
2/19/2016
18
35
0
10
20
30
40
50
60
70
80
90
100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Design Production Ramp-up Volume
New YieldLearningCurve
YieldLearning Curve
Design Yield Optimization
Fab Yield Optimization
Yield Life Cycle CurveYield Life Cycle Curve
Yie
ld A
sses
smen
t (%
)
36
0
10
20
30
40
50
60
70
80
90
100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Design Production Ramp-up Volume
New YieldLearningCurve
YieldLearning Curve
Design Yield Optimization
Fab Yield Optimization
Yield Life Cycle CurveYield Life Cycle Curve
Yie
ld A
sses
smen
t (%
)
2/19/2016
19
37
0
10
20
30
40
50
60
70
80
90
100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Design Production Ramp-up Volume
New YieldLearningCurve
YieldLearning Curve
Design Yield Optimization
Fab Yield Optimization
Yield Life Cycle CurveYield Life Cycle Curve
Yie
ld A
sses
smen
t (%
)
38
Expect Many Types of “Things” Highly Fragmented Market
• By 2016, 50% of Internet of Things solutions will originate in startups less than three years old.
– Expect 10 billion shipments in 2020
– Many smart versions of existing product markets
– Key challenge: where to focus?
Source: Gartner, Preliminary, Sep 2013
2/19/2016
20
39
Significant Growth in SensorsSensors Are Everywhere
Sensor Units to Grow to 30 Billion Units in 2017
0
5000
10000
15000
20000
25000
30000
2009 2010 2011 2012 2013 2014 2015 2016 2017
Consumer
Automotive
Computing
Smartphones
Feature Phones
Industrial
Other Communications
Source: Semico Research, 2013
30B Units
10B Units
40
How the Billions of “Things” are Connected…
IoT Edge DevicesAggregation Layers
(Hubs/Gateways)Remote Processing
(Cloud Based)
“Things” with sensors & actuators that monitor
and control
Connectivity & Interfaces to aggregate the edge data to
send to the cloud
Applications to analyze the data and offer cloud
services
2/19/2016
21
41
The IoT Opportunity GapThe IoT Opportunity is much largerAnalysts predictions forconnected devices (2020):
30 billion?50 billion?75 billion?
Rea
ch
Time
The IoT Market is growingNot new concept , it’s been around for >20 years1
Connected things > world population (6.8B)
Today
Silos of Things
1 Weiser, Mark (1991) “the Computer for the 21st Century”The term Internet of Things was proposed by Kevin Ashton in 1998
1 Weiser, Mark (1991) “the Computer for the 21st Century”The term Internet of Things was proposed by Kevin Ashton in 1998
42
Relationship: Users, Devices & ServicesFunctional Becomes IOT Data /Leveraging data enabled services revolution
Functional Data
2/19/2016
22
43
IoT Edge DevicesA Market Segmentation
SmartAppliances
Smart CitiesMetering
Safety &Security
Commerce
WearableInfotainment
Health &Fitness
Wea
rab
leD
evic
esM
ach
ine
to M
ach
ine
44
Wearable Devices
Google Glass 570mAh Li Polymer 1 days
Samsung Gear S Smart Watch 300mAh Li-Ion 2 days
Samsung Gear Fit Smart Watch 210mAh Li Polymer 4-5 days
Starkey Hearing Aid 91 - 630mAh Zinc Air 3-22 days
source: IEEE CS & ComSoc Joseph A. Paradiso, Thad Starner 2005
Laptop Computing Technology Improvements
2/19/2016
23
45
A Better Source Of PowerWe Have Potential!
Implanted Glucose Cells + Body Heat
Shoe Insert + Walking Motion
source: Joseph A. Paradiso, Massachusetts Institute of Technology Media Lab,Thad Starner, Georgia Institute of Technology, GVU Center
46
Bodies In Motion…
Walking the streetsEnergy harvested: 1 square generates up to 2.1 watts
Stepping in your shoesEnergy harvested 7- 67W possible but practically around 1W
Wearing clothesEnergy harvested: uW
Shaking that thingEnergy harvested: 5-12 W
source: postscapes.com
2/19/2016
24
47
Implanted Device RF/Thermal/Piezo Powered
Ear implants
• Energy requirements– active 500 μs per minute ~ 60 μJ
• Energy content of an implantable Li-ion battery is ~200mAh (Quallion)– would last 1*E+5 hrs = 11.4 years!!
48
Implanted Device RF/Thermal/Piezo Powered
• Exploring MIM Cap capabilities
• MIM Cap: 38 fF / μm2– 4 mm x 4 mm -> C = 0.6 μF
– Total Energy = 2.7 μJ @ 3V
• Energy requirements– active 500 μs per minute ~ 60 μJ
– NOT feasible with MIM
– BUT easy with SuperCaps
• Harvesting Piezoelectric (blood pressure) or thermal energy– Blood pressure @.37W can easily
sustain energy needed
2/19/2016
25
49
Wearable Systems Thermally Powered
Source: Skinny player URL. Designers: Chih-Wei Wang and Shou-His Fu http://www.energyharvestingjournal.com/articles/body-heat-powered-music-player-00002892.asp?sessionid=1
• Low power embedded processors, along with innovative energy harvesting and storage technologies will make many autonomous, connected "Things" possible
50
Table of Contents
6%
15%
17%
29%
33%
0% 20% 40%
Other
Smartglass
Activity Tracker (Lifelog)
Smart/MobileHeadset/Headphone
Smartwatch
Portable Medical Devices
Wearable Devices/Fitness/Portable Medical Devices
20146%
6%
9%
11%
13%
19%
35%
0% 10% 20% 30% 40%
Other
Smart City, such asDigital Lighting
Machine-to-Machine
Smartmeter
EmbeddedVision/Sensor
Smart Home/ SmartAppliances
WearableDevices/Fitness/Portable
Medical Devices
Internet of Things
2014
What is the PRIMARY application of your design?GLOBAL 2014
Synopsys Global User Survey 2014
2014 N = 161 2014 N = 52
No historical data as these are new questions starting 2014
Required– WS_BYPASS – Allows normal (functional) mode and puts the wrapper into bypass mode. – WS_INTEST – Allows internal testing using a single chain configuration in the WBR.– WS_EXTEST – Allows external test using a single chain configuration in the WBR.Optional– WP_EXTEST – Allows external test using a multiple scan chain configuration in the WBR.– WS_SAFE – Puts the core into a quiet mode and outputs a predefined static (safe) state from
all output ports. It also puts the WBR into bypass mode.– WS_CLAMP – Outputs a programmable static (safe) state from all output ports. It also puts
the wrapper into bypass mode.– WS_PRELOAD – Loads data into the single silent shift path of the WBR.– WP_PRELOAD – Loads data into the multiple silent shift path of the WBR.– WS_INTEST_SCAN – Allows internal testing by concatenation of the wrapper chain with a
single internal chain.• Specific Instructions
– WS_WBR_SEL - Selects wrapper boundary registers between WSI and WSO.– WH_DIRECT - Configures the cells to use parallel test inputs/outputs during a test.– WS_MODE_SEL – Configures the configuration of the cell.– WS_STATIC_SEL - Selects the static register for forcing values to user pins.– WS_STATIC – Configures the cells to force the content of the static register to corresponding
user pins.– WS_SCAN_SHIFT - Configures the cells to be automatically updated during shift.– WS_ID_SEL – Selects the unique ID of the wrapper.– WS_<CORE_CHAIN_ID>_SEL – Selects the core’s test chain between WSI and WSO.
88
Two Compliance Levels
• IEEE 1500 Prepared– Core does not have a complete IEEE 1500 wrapper function
– Core has a complete IEEE Information Model, which accurately describes the core’s tests, as well as provide all information on the basis of which the core could be made ‘IEEE 1500 Wrapped’ (either manually or automatically by tools)