Dilip Gopi and Roland Schelasin Industrial Engineering, Texas Instruments, Maine

Outline Wafer Fabrication Process Overview TI Maine Fab Overview Scheduling Methodology Overview Problem Statement Rule Definitions Benefits 2 © 2014 Texas Instruments Incorporated

TI Maine Fab Overview 3 © 2014 Texas Instruments Incorporated

Wafer Fabrication Process Overview 4 Clifford L. Henderson Used with Permission from Clifford L. Henderson, GA Tech 404 Wafer Fab Re-Entrant Flow © 2014 Texas Instruments Incorporated

TI Maine – Location Freising FFAB Dallas DFAB DMOS5 DMOS6 SFAB DHC RFAB DBump Aguascalientes TMX Baguio Malaysia TIM TIEM Taipei TITL Miho Miho Fab Hiji Hiji Fab Hiji A/T Stafford HFAB HBump Color Management Fab Assembly & Test Bump Pampanga Clark Aizu Aizu Fab Chengdu Chengdu Fab Greenock GFAB Portland MFAB © 2014 Texas Instruments Incorporated

MFAB – Maine – The Way Life Should Be! Portland © 2014 Texas Instruments Incorporated

TI Maine - Fab Overview 7 200mm SMIF w/Intercell Transport/Stockers 27 Technologies, nm, Masks 50,000 Reticles (10,000 In Fab) 700 Active Products 75k sq. ft. Clean Room In-House MES 7 x 24 Manufacturing with 4 Shifts Electronic Run Cards (STS): Real Time Wafer Level Tracking Intelligent Stockers for Lot Dispatch Automatic Recipe Select/Downloading Automated “Feed Forward” Adjustment for Advanced Process Control © 2014 Texas Instruments Incorporated

TI Maine – Mix and Capacity Overview Approximate Wafer Starts Mix 54% 350nm CMOS, Avg. 26 Patterns 26% nm BiCMOS, Patterns 20% 180nm CMOS, Avg. 31 Patterns Bottle Necks Design – Photo Tactical– EPI, Etch (Metal, Oxide), CMP, Diffusion (Clox) – Technology + Process Module Sub-Capacities 8 © 2014 Texas Instruments Incorporated

Scheduling Overview Scheduling is an important tool for manufacturing and engineering since it can have a major impact on the productivity of a process. Helps minimize –Production Time –Costs by explicitly telling the manufacturing specialists what to run on which equipment. Production scheduling aims to maximize efficiency of the operation The methodology presented uses scheduling theory based on point allocation to dictate lot-to-entity allocation and visualization for one of the constraint modules in the factory 9 © 2014 Texas Instruments Incorporated

Problem Statement 6 Chamber States  (setup condition) with 35 active production recipes In-flexibility to run everywhere based on, –Hardware –Gas restrictions –Recipe qualification on tools Chamber conditioning & test runs needed when switching states Inability to meet required productive hours results in cycle time losses  Chamber states are defined based on the gas configuration used for Production © 2014 Texas Instruments Incorporated

Problem Statement (Contd.) Goal Productive Hours/Day Components of Chamber State Change Re-Measure (  6 Months Data) Equivalent Loss of 1 Tool due to © 2014 Texas Instruments Incorporated

Proposed Model – Tools by State Reduce Chamber State Changes Eliminate associated Idle time with State Changes Use Passing Lanes to keep recipes Qualified Each Tool to have a Primary and Backup State EPI Scheduler to define State Changes based on WIP and Look ahead functions Reduce downtime associated with state changes  © 2014 Texas Instruments Incorporated

What is a WIP Scheduler? The WIP Scheduler is used to schedule a lot to be loaded into the tool The WIP Scheduler uses a set of rules that have been set up for each entity type The WIP Scheduler schedules a run by using the rules and other factors on each tool The WIP Scheduler and Dispatch are Different! –Dispatch only shows work, and the order is based on dispatching rules: Slack time Priority lot (Hot, Hand carry, etc.) –The WIP Scheduler sees what can run on a tool and determines what lot or group of lots (batch) should run next based on the rules 13 © 2014 Texas Instruments Incorporated

Rule Definitions Points are assigned for each rule (Low, Med, High) Lot-based Rules –Priority, Lateness, Static Lot / Entity Rules –“Cost” of changing Chamber State –How many tools can run the Recipe right now? Fewer tools means MORE points –MTTD consideration for batches with inspectable lots © 2014 Texas Instruments Incorporated

Maximization Function Where Rules are defined based on, Lot Priority Lot Lateness Lot Slack Time Lot Inspection Tool Chamber State Tool Count Tool Ready Tool State, etc. 15 © 2014 Texas Instruments Incorporated

EPI Rule Set TypeName Min Points Mid Points Max Points Min TextMid TextMax Text BATCHPRIORITY no priorityMedium PriorityHigh Priority Lot BATCHLATENESS days late7 days late >10 days late or push it BATCHSTATIC hours48-72 hours>72 hours BATCHSTATIC_HRS001hours static BATCHINSPECTION Inspectable Inspectable Inspectable >500 ENTITYTOOL_COUNT tools2 tools1 tool ENTITY QUAL EXPIRATION >7-14 days2-7 days<2 days ENTITYRUNNABLE Cannot Run Allowed but must Qual Can run Now ENTITYSTATE_CHANGE000Value from table © 2014 Texas Instruments Incorporated

State Change Table To From StateAs HiAs LoBaseBoronP HiP Lo As Hi As Lo0 100 Base Boron0100 P Hi P Lo Risk Levels 0Just do it. 100Can be handled with standard prep recipe (hours) 1000Extended chamber prep time and qualification (days) 10000Don't do it. © 2014 Texas Instruments Incorporated

Scheduler Breakdown Projected time when tool will finish current lot(s) Current State Points Score based on Rule Set © 2014 Texas Instruments Incorporated

Benefits 34% increase in Productive Hours Chamber State Dedication Helped, –Increased Availability –Increased Utilization –Increased Labor Productivity –Reduced number of State Changes –Reduced Quals –Gas Savings 19 Scheduler Online © 2014 Texas Instruments Incorporated

Benefits (Contd.) Empowerment to the shop floor Engineering Productivity Improvement Gave the floor a better visibility of what needed to run next Look ahead of when the next lots are needed on what tool 20 © 2014 Texas Instruments Incorporated

Q&A 21 © 2014 Texas Instruments Incorporated