The document discusses the benefits of using SEMI's Equipment Data Acquisition (EDA) standards for smart manufacturing applications. It describes how EDA standards can help factories collect the right equipment data to develop multivariate fault detection models, which use real-time sensor data to detect process drift and tool failures. The metadata models in EDA Freeze II and EDA Common Metadata standards provide context to select the proper fault detection algorithms and allow data collection plans to be changed flexibly as fault models evolve.

1. 18th European Advanced Process Control
and Manufacturing (apc|m) Conference
Dresden, Germany
Alan Weber – Cimetrix Incorporated
EDA* Applications and Benefits
for Smart Manufacturing
* EDA = SEMI’s Equipment Data Acquisition standards suite

2.  Key messages
 Keeping score with ROI
 The standardization paradox
 Data collection strategy
 EDA factory applications
 Example use case
 Factory implementation alternatives
Outline

3.  It’s important to keep score… while agreeing on the
rules of the game
 Standardization does not lead to conformity, but
actually enables genuine transformation
 Data collection should be viewed as a spectrum of
alternatives, not an all-or-nothing proposition
 Fabs collect data to solve specific problems… not to
make life difficult for their equipment suppliers
 The latest generations of SEMI standards offer many
opportunities for engineering cost savings
Key messages

4. The importance of keeping score
It’s an integral part of everyday life
 Sports
 Business
 Politics
 Science
 Health
 Finances
 Relationships
 …

5. Keeping score with an ROI model
Agree on relative cost and value of key factors
Costs
 Materials/outside services
 Software development
 Technology development
 Hardware
 Licenses
 Internal labor
 Operations
 Engineering
 Automation
 Information technology
 Capital expenses
 Equipment
 Other
Benefits
• Product material
• Yield
• Yield ramp
• Scrap reduction
• Time
• Equipment/fab uptime
• Factory cycle time
• New Product Introduction time
• Cost Reduction
• Qual wafers
• Hardware
• Licenses
• Engineering labor
• Other

6.  How can standardization result in competitive advantage?
 If all fabs can gather the same data, how can a single fab
differentiate itself from the others?
 Consider the use of time
 We all start with the same amount of time... but success favors
those with the discipline to use it wisely
 Think of discipline as a sort of “personal entropy reduction”
 The same applies to data
 It’s not what you collect, but how you use it
The standardization paradox
Avoiding the conformity trap

7. Standardization
conformity…
Transformation !

8.  What are the critical decisions that
need to be made?
 What insights do we need to make
these decisions?
 What information do we need to get
these insights?
 What data must we collect to get the
information we want?
Data collection strategy
Start with the goal and work backwards
Source: McKinsey & Company
Insights
Information
Data
Decisions

9. KPIs, stakeholders, applications, …
Importance of the equipment model
9
Key
Performance
Indicators
(KPIs)
Factory
Stakeholders
Manufacturing
Applications
Data
Equipment
Model

10. EDA purchase spec development
Process and results
10
Factory
Stakeholder
Questionnaire
Generic EDA
Purchase Spec
Outline
Stakeholder
Answers
Factory
EDA Purchase
Specifications
Manufacturing
Technology
Development
Process-specific
Supplier
Response Forms
Automation and
Integration
Supplier
Relations
ApplicationsInfrastructure

11.  Industrial engineers
 Responsible for monitoring equipment and factory throughput in real-time, identifying opportunities to eliminate wait time
waste in individual equipment types as well as the overall factory, and addressing bottlenecks as they shift and emerge;
 Production control staff
 Responsible for determining the material release schedule and managing the factory scheduling/dispatching systems to
accommodate changes in customer orders and/or factory status;
 Equipment engineers
 Responsible for fleet matching and management to minimize or eliminate the need to dedicate certain equipment sets for
critical process steps and thereby simplify the overall factory scheduling process;
 Maintenance engineers
 Responsible for minimizing equipment downtime, MTTR (mean time to repair), and test wafer usage required to bring
equipment back to production-ready state;
 Facilities engineers
 Responsible for collecting and integrating sub-fab data from pumps, chillers, exhaust systems, and other complex
subsystems into the production data management infrastructure for use by a growing range of analysis applications;
 Sensor integration specialists
 Responsible for supplementing the built-in sensing and control capabilities of critical process and measurement equipment to
support advanced process development…
 Program management
 Responsible to executive management for competitive manufacturing results
Factory stakeholders
and their careabouts
11

12. Data collection alternatives
Spec contents, tool capability
SEMI Standard Level Functionality Benefit
GEM/GEM300 Full support for E40, E87, E90, E94, etc.
Baseline:
Supplier-specific integration costs; labor-intensive SECS data
collection management, tool characterization, software upgrade
verification, and fault model development processes
EDA Freeze I
(1105)
EDA basics – early metadata models, DCP-based
“data on demand”, multi-client access
Self-documenting interface capability; quick and easy to change
data collection plans as application needs evolve; factory system
architecture flexibility
EDA Freeze II
(0710)
Conditional triggers in trace requests, simple event
support, interface discovery; second-generation
metadata models
Precisely “frame” trace data depending on application
requirements; one-click connectivity; cleaner model structures with
richer event/parameter content; higher performance
EDA Common Metadata
(E164)
Complete coverage of GEM300 and E157 objects,
state machines, events; standard metadata model
structure, content, and names
Programmatically generate DCPs, configure generic tool
applications, characterize equipment behavior; simplify mapping
to factory data management systems
Factory-Specific
EDA Requirements
Process-specific parameters for advanced feature
extraction for FDC, PHM, VM; mechanism- and
component-level command/response signals for
fingerprinting, tool matching; etc.
Dramatically increase visibility into tool and process behavior;
enable advanced “smart factory” monitoring and control
applications well beyond current capabilities

13.  E120/E125 Common Equipment Model usage/content
 Nodes and parameters must have meaningful descriptions
 Equipment element attributes for all E120 nodes must have meaningful
values
 All definitions (exceptions, SMs, parameter types, units, SEMI object
types) must be referenced
 Strict event name enforcement
 State Machines
 Strict State Machine definitions
 Requires E157 State Machines for all process modules
 Requires E90 State Machines for all substrate locations
 Requires all Parameters, Events and Exceptions defined in Freeze II
standards to be present
 State and transition names must match GEM300 standards
What does E164 specify?
Structure and content of equipment metadata

14.  Model structure exactly reflects tool
hardware organization
 Complete description of all potentially
useful information in the tool
 Always accurate, always available – no
additional documentation required
 Common point of reference among tool,
process, and factory stakeholders
 Source of unambiguous identifiers/tags
for database [auto] configuration
 Enables “plug and play” applications
Standard metadata model benefits
First specified by SEMI E164
Process Module #1
Gate Valve Data
Substrate Location
Utilization
More Data,
Events, Alarms
Process Tracking
Other
Components

15.  Key messages
 Keeping score with ROI
 The standardization paradox
 Data collection strategy
 EDA factory applications
 Example use case
 Factory implementation alternatives
Outline

16.  Real-time throughput monitoring
 Precision FDC feature extraction
 Specialty sensor data access
 Fleet matching and management
 eOCAP execution support
 Sub-fab data integration/analysis
 Automated equipment characterization
EDA factory applications
Current leading edge
16
Wide range of stakeholder coverage

17.  Description
 Multivariate statistics used to develop reduced-dimension
equipment fault models for various operating points
 Fault models evaluated in real-time to detect process
drift and/or impending tool failure
 May interdict tool operation in mid-run to prevent/reduce
scrap
 KPIs affected
 Process yield and scrap rate through higher detection
sensitivity
 Equipment availability resulting from fewer false
positives
 EDA leverage
 Conditional triggers in trace request frame data collection
 Metadata model contains context to select proper fault
detection algorithms
 Multi-client access to develop, evaluate, and update fault
models
Example use case
Multivariate Fault Detection and Classification (FDC)

18. Data collection alternatives
Fault Detection and Classification (FDC)
SEMI Standard Level Functionality Benefit
GEM/GEM300
Fault models difficult to change after initial development if data collection
requirements change
Baseline
EDA Freeze I
(1105)
Easy to change equipment data collection plans as fault models evolve
and require new data;
Model development environment can be separate from production system
Engineering labor reduction; improved fault
models and lower false alarm rate
EDA Freeze II
(0710)
Use conditional triggers to precisely “frame” trace data while reducing
overall data collection needs; Incorporate sub-fab component/subsystem
data into fault models
Even better fault models; reduced MTTD
(mean time to detect) of fault or process
excursion; little or no data post-processing
required
EDA Common Metadata
(E164)
Include standard recipe step-level transition events for highly targeted
trace data collection;
Automate initial equipment characterization process by using metadata
model to generate required data collection plans
Faster tool characterization and fault model
development time
Factory-Specific
EDA Requirements
Incorporate previously unavailable equipment signals in fault models;
Update data collection plans and fault models automatically after process
and recipe changes;
Include recipe setpoints in the equipment metadata models
TBD (Not yet applicable)

19.  Factor values
 Number of tools - 2000
 Hour of tool time - $2200 (average raw and finished wafer value)
 Qual wafer cost - $250
 Hour of engineering/tech time - $150
 Cost of false alarms
 Tool time to resolve (incl. 0.5 hour metrology) – 5 hours
 Qual wafers required – 6
 Engineering/tech time required – 2 hours
 Cost per false alarm = 4.5*2200 + 6*250 + 2*150 = $11,700
 False alarm rate – 2 per tool per year
 Total false alarm cost = $11,700*2000*2 = $46.80M
 Benefit of advanced data collection
 Reduction in false alarm rate – 50%
 Annual savings = $23.4M
ROI factors and FDC false alarm costs
Hypothetical megafab

20.  Factor values
 Wafer value - $10,000 (average cost of WIP)
 Hour of engineering/tech time - $150
 Cost of process excursions
 Wafers per excursion – 500
 Delta yield per excursion – 3%
 Engineering time required to resolve – 160 hours
 Cost per excursion = 500*10,000*.03 + 160*150 = $174,000
 Excursion rate – 24 per year
 Total excursion cost = $174,000*24 = $4.12M
 Benefit of advanced data collection
 Reduction in # and severity (yield loss) of process excursions – 25%
 Annual savings = $1.72M
ROI factors and process excursion costs
Hypothetical megafab

21.  Key messages
 Keeping score with ROI
 The standardization paradox
 Data collection strategy
 EDA factory applications
 Example use case
 Factory implementation alternatives
Outline

22. Factory Architecture A
Application-driven multi-client EDA connections
CVD
Furnace
CMP
GEM300
EI
GEM300
EI
Tool
Etch
MES
YMS
APC, RMS
HTTP
Factory Information
and Control Systems
Process
Data
Storage
HSMS
AMHS
SOA
MES,
RMS
Other DB
EDA
Applications
(EES, PCS)
Application
Data
HTTP
HTTP

23. Architecture style
Wild West – chaotic

24. Factory Architecture B
Add-on fab-wide EDA infrastructure
CVD
Furnace
CMP
EDA
EI
GEM300
EI
GEM300
EI
Tool
Etch
MES
YMS
APC, RMS
HTTP
Factory Information
and Control Systems
Equipment
Integration
Servers
(Enhanced)
Process
Data
Storage
HSMS
AMHS
EDA/GEM
Time-Series
DB
SOA
EDA
Admin
EDA
Config
Context
data MES,
RMS
Other DB

25. Architecture style
Evolutionary

26. Factory Architecture C
Integrated production system architecture
26
GEM/EDA
Gateway
HSMS / HTTP
Equipment
Gateways
GEM/EDA
Gateway
SOA
MES
YMS
APC, RMS
Factory Information
and Control Systems
Process
Data
Storage
AMHS
EDA/GEM
Time-Series
DB
Equipment
Data Mgmt
Support
Data Mgmt
Config
MES,
RMS,
Other DB
CMP
Furnace
Tool
Etch
CVD
New Mfg
Applications

27. Architecture style
Classical

28.  감사합니다
 唔該
 Merci
 Danke
 多謝
 ありがとうございます
 Gracias
Acknowledgements and Thanks
 Mark Reath, Member, Technical Staff, GLOBALFOUNDRIES
 Decades of apc|m Conference organizers and participants!