Interoperability Beyond Design

1
Interoperability Beyond Design
Sharing Knowledge between Design and
Manufacturing Don Cottrell VP Emerging
Technologies Si2 Corporation Thomas J.
Grebinski SEMI Data Path Task Force Chair SEMI
Universal Data Model Working Group Chair
2
Data-centric IC Designs

• The IC design community is becoming more data
  centric.
• Overcoming the issues of interoperational
  efficiencies.
• Common data structures for the entire IC design
  data flow.
• Open data structure standards
• OpenAccess and Si2
• Open-source applications program interface.
• Open-source reference data base for all IC design
  data.
• Free license for anyone to use and redistribute..
• The realization of common expectations
• within an increasingly complex, more intimately
  connected and data-intensive design data flow.

3
Open-source Reference Data Base and API

• OpenAccess Coalition and Si2
• Click-thru internal use license
• Free access to API specification
• Free access to reference database binaries
• Signed internal use and distribution license
• Free access to reference database source
• Royalty-free redistribution rights (binaries
  only)
• Rights to distribute modifications to the
  reference database (binaries only) that do not
  alter the API

4
OpenAcess Design Data Base and API
Commercial EDA Tools
University Research
Internal Proprietary Tools External Design
Partners
Standard API
The realization of common expectations
.
Reference Database Cores/Cells Gates Transistors
Physical Layout OASIS (planned)/GDSII
5
Wide range of Functionality

• Design Database
• Chips / Blocks / Cells
• RTL to Silicon
• Digital / Analog
• Automated / Custom
• Logical / Physical
• Batch / Interactive
• Library Database
• Design organization
• Design management
• Access control
• Technology Database
• Foundry Rules
• Design Constraints

• Translators
• GDS2
• OASIS (planned)
• LEF
• DEF
• Verilog (planned)
• SPEF (planned)
• SPICE (planned)

6
Opaque View Beyond IC Design

• The data view beyond IC design is opaque after
  GDSII and OASIS.
• Design intent is not conserved beyond GDSII and
  OASIS.
• GDSII and OASIS are a geometric encapsulation of
  the design.
• SEMI P10 and text files fed downstream are
  another encapsulation (interpretation) of the
  design intent.
• The basic design data structure has changed
  substantially and thus, the design and mask
  engineering view up and downstream is not
  immediately accessible and also not clear
  (unambiguous).
• Different data structures between design and mask
  manufacturing centers.
• Obscures the up and downstream view further.
• Need to convert to standard record structures for
  mask manufacturing.

7
IC Design Data Encapsulation for Mask
Manufacturing
Conversion to standard record types
SEMI P10 Text Files
Physical Layout
Internet File Transfers
Cells, Cores, Gates, Transistors
Mask Technical Planning Data Base
Design Data Base
OASIS GDSII
Geometric Shapes
OASIS/GDSII
8
Mask Data Preparation and Storage
MIC, MEBES, Toshiba Hitachi, JEOL, KLA, etc.
Data Fracture
Tonality, Mirroring
Data Finishing
Sizing, Biasing,
Mask Layout
Frame generation, bar coding, test structures
Layer Extractions, Scaling, Shrinking
Mask Technical planning Data Base
OASIS/GDSII Mask Layout Data
Data Preparation
9
Machine-specific Data Flow
Pattern Generation
MIC, MEBES, Toshiba Hitachi, JEOL, KLA, etc.
Job Queue Data base
Mask Technical planning Data Base
OASIS/GDSII Mask Layout Data
Mask Inspection
Mask Critical Dimensions
10
Data Flow and Transformation in a Mask Writer
Rendering
Rendering
Extraction
CFRAC
Writer Job Queue Data Base
Mask Pattern Generation
CFRAC
FRAC
Extraction
Viewers- Integrity Test
11
Changes in Structure and Hierarchy of Design and
Manufacturing Data
Encapsulation Filter
Record Types Filter
Data preparation Filter
Technical Planning Data Base
Technical Planning Data Base
12
The Fracturing of Fractured Mask Layout Data
Rendering
FRAC
CFRAC, Extraction and Rendering Filter
Writer Job Queue Data Base
Technical Planning Data Base
Mask Pattern Generation
CFRAC
FRAC
Extraction
13
Design and Manufacturing Interoperability

• IC and mask design intent are lost downstream
• Numerous filters and translations of data drop
  design and manufacturing hierarchy.
• There is no real time direct data link to what is
  written on a mask.
• Which gives the IC designer an opaque view of how
  the design is spatially transformed onto a mask.
• Such a view is a compelling need when
  non-parasitic biasing and proximity corrections
  push a design outside acceptable IC design rules.
• Such constraints will increase as the spatial
  density budget for circuit elements increases.
• Such losses extends time to market and increase
  design and manufacturing costs.
• It will become increasingly important to design
  at a higher level of abstraction with a real-time
  view of what is actually writable on a mask or
  wafer.
• It will also become increasingly important to
  ensure that any movement of data is as efficient,
  portable and extensible as possible.

14
Finer Data Granularity Expected at the Mask
Pattern Generation Level

• Design and manufacturing data granularity
  continues to increase.
• 100 to 300 Gb files with the expectation that the
  cost of ownership to write or manufacture a mask
  does not increase.
• Job Queue and sorting data rates are reaching
  several hundreds of Mb/second with needed storage
  capacity in the Terabytes region.
• These types of demands run orthogonal to the
  necessary reduction of COO and a shorter time to
  market.
• Greater access to design data
• Interconnectivity data will help improve mask
  layout and engineering data processing time.
• Massive parallel processing
• Sorting is fundamental for efficient massive
  parallel processing
• Access efficiency features, such cell reference
  tables and bounding boxes, accelerates
  transformations, such as sorting

15
The SEMI and OpenAccess Universal Data Model (UDM)
Standard API
.
IC Design and Mask Technical Planning Data Base
Hierarchical Data Base
Cells, Cores, Gates, Transistors
Physical Design Layout
Geometric shapes
Mask layout
Write-, inspection- and CD measurement-ready data
files
16
The SEMI and OpenAcess UDM

• One standard data base and applications program
  interface (API).
• For design and mask manufacturing.
• Access to all design and mask manufacturing data
  through one common data language.
• Standard semantics for all model objects,
  attributes and relationships
• No ambiguity between or within design and
  manufacturing tasks.
• Hierarchy preserved through to mask pattern
  generation, CD measurements and inspection.
• Unified design to manufacturing data flow and
  effort.

17
The SEMI and OpenAcess UDM

• Full access to design and manufacturing intent up
  and down the design and manufacturing flow.
• Opens the door to more efficient massive parallel
  processing at the mask pattern generation level.
• Thread safe and data preparation
• multiple threads within an application can
  operate on data in parallel without the risk of
  one thread contaminating the data on another.
• request objects that are within a specified
  spatial area (e.g. a stripe)
• Query by Region for pattern generation and
  inspection
• multiple stripes in parallel multiple
  executions.
• Direct access to mask layout and DRC extraction
  data by region of interest during mask inspection
  and CD measurement operations.
• Unlimited scalability
• The model is limited only by the ability to store
  and manage the data within the data base.

18
Extending the reach of the UDM
.
Standard API
Universal Data Model
Hierarchical Data Base
Wafer process data
Process Parasitics
19
UDM Status

• The technology exists today for full
  implementation
• The reference data base and API are already
  available and being used today.
• Mask and wafer implementations underway through
  SEMI, OpenAccess Coalition and Si2.
• It is a community resource
• Change-order team in place.
• Applicable to design, mask and wafer data flows.
• Applicable to high-speed and volume data
  rendering and measurement applications.
• Free-use license
• Free-use after release from the OpenAccess
  coalition.

20
Organizations Involved

• SEMI Data Path Task Force
• Tom Grebinski (tgrebinski_at_sbcglobal.net)
• Task Force Chair
• Applicable site
• www.semi.org
• www.si2.org/eda-mask
• SEMI UDM Working Group
• Tom Grebinski and Don Cottrell
• Working Group Co-chairs
• OpenAccess Coalition and Si2
• Scott Peterson, LSI Logic (OAC Chairman)
• Steve Schulz- President and CEO, Si2
  (schulz_at_si2.org)
• Applicable sites
• www.openeda.org
• www.si2.org /openaccess

21
Companies Participating

• Cadence Design
• Mentor Graphics
• Synopsys
• Micronic Laser Systems
• JEOL
• KLA-Tencor
• Applied Materials
• Dai Nippon Printing
• Toppan Printing
• Photronics
• Dupont Photomasks
• STMicroelectronics
• TSMC USA
• Toshiba (NuFlare)
• Hitachi
• Hewlett-Packard

• International Sematech
• IBM
• Infineon
• Texas Instruments
• Motorola
• Philips Semiconductor
• JEOL
• Alcatel
• Intel
• AMD
• SELETE/JEITA
• National Semiconductor
• LSI Logic

22
Moving Forward

• Critical Path Items
• New members to SEMI Data Path Task Force and
  Working Groups
• New members to the OpenAcess Coalition
• Implementation of the mask and wafer extensions
  to the OpenAccess Data base and API.
• Extending and then bridging the responsibilities
  between design, mask and wafer manufacturing.
• Data-intensive flow integration with OpenAccess
  and the UDM.
• Pattern generation
• Data Preparation
• Mask inspection and CD measurement
• Formalize relationship between SEMI, SI2 and
  OpenAccess Coalition

23
Successful Track Record

• Si2 and the OpenAccess Coalition
• Open-Source Reference Data Base and API
• Source and Binary code made available for the
  model and API.
• Unprecedented effort and availability
• Released to the public January 1, 2003
• SEMI Data Path Task Force
• The development of a replacement for GDSII called
  OASISTM.

24
Successful Track Record

• OASISTM
• 64-bit Open Artwork System Interchange Standard.
• vs. 16-32 bit
• direct access to cell pointers
• GDSII has only sequential access to cell data.
• 10-50 times more compact
• Makes use of modality
• Can mimic data organization of virtually any
  writing or inspection pattern file. 
• Flexible property mechanism which can be used to
  tag figures, arrays, and cells with as much
  textual and numeric information as needed by
  downstream processors.

25
Successful Track Record

• The creation of the Universal Data Model (UDM)
• Embrace of the technology worldwide and across
  several industry platforms.
• A recognition of the importance of the link
  between design and manufacturing by the industry
  and the media.
• Adoption of the OpenAccess reference data model
  and API as the basis for the ongoing development
  of the UDM.

26
Adoption of the Technology

• The compelling need
• Data granularity of an IC design and the
  manufacturing of an IC continues to increase.
• Closer integration of the data flow greater
  opportunity to lose design intent at a number of
  levels.
• Greater inability with the tracking and credible
  use of design and manufacturing data.
• Greater need for speed with fewer errors.
• Massive parallel processes into the tera-pixels
  per second.
• Greater need for deign intent at the mask data
  preparation, pattern generation and inspection
  levels.
• Greater need for manufacturing intent in the
  design space cross talk is there ad getting
  louder.
• Proximity corrections at the manufacturing level
  render DRC at the design level less effective.
  The need for a view by design at the mask
  production level.
• There is no choice