How to Successfully Transfer MEMS from a University Lab to a Commercial Foundry Transducers 2015, Anchorage, AK Alissa M. Fitzgerald, Ph.D. | 23 June 2015
12th
anniversary
Overview • • • •
About AMFitzgerald Are you ready? Choosing a foundry Transfer
Page 2 Transducers 2015
© AMFitzgerald 2015
AMFitzgerald services
MEMS Innovation
MEMS Solutions
Technology Strategy
Creation of novel designs and IP
Paths to manufacturing and market
Key insights from MEMS experts
Page 3 © AMFitzgerald 2015
Full development services from concept to production
AMFitzgerald in‐house
Strategic partners
• Advantages of working with us: – Multi-disciplinary, expert engineering team focused on MEMS development for volume production – Rapid prototyping on state-of-the-art tools common to foundries – Own all the design and process IP – Bring a mature, de-risked design to the foundry to get better pricing and faster time to production
Page 4 © AMFitzgerald 2015
Are you ready?
Are you ready to go to foundry? • YES, if you have: – Stable process flow and mask set – Budget (> $1M/yr.) – Order schedule for next 1-2 years (i.e. customers) – Cost targets NewCo’s Production Order Schedule
Production Number of Wafers
Target $/Wafer
Year 1
500
$1500
Year 2
1000
$900
Year 3
1500
$700
Page 6 Transducers 2015
© AMFitzgerald 2015
Are you ready to go to foundry? • NO, if you: – Are still exploring the physics of your devices and trying to improve them significantly – Need Design of Experiments to characterize your device behavior (i.e. many design variants) – Don’t yet fully understand what design/process conditions create a “good” device – Have not yet secured > $1M in funding (just for MEMS fabrication)
Page 7 Transducers 2015
© AMFitzgerald 2015
Do you need other manufacturing partners? Assembling an ecosystem takes time and resources
MEMS Product Company
Page 8 Transducers 2015
© AMFitzgerald 2015
Viable business models for new MEMS companies “Real men have fabs.” – Jerry Sanders, AMD Today: “Smart MEMS companies don’t have fabs” • Fabless – Plenty of capacity at 150 and 200 mm • Hybrid or Fab-Lite – Adds value with specialty processes/equipment that are kept in-house
Page 9 Transducers 2015
© AMFitzgerald 2015
Read the MIG Foundry Engagement Guide • Written by a group of foundries and MIG members • Practical, how-to information http://memsfoundry.wikia.com/wiki/MEMS_Foundry_Engag ement_Guide_Wiki#Background_Information
Page 10 Transducers 2015
© AMFitzgerald 2015
Largest MEMS Foundries
Source: Yole Developpment, 2015 Page 11 Transducers 2015
© AMFitzgerald 2015
Foundries for smaller wafer orders (< 100 wafers/year) • • • • • • • •
Micrel (US) MEMSCAP (US) MidWest MicroDevices (US) LioniX (NL) Micronit (NL) MEMS Core (JP) INEX (UK) CSEM (CH)
• Larger foundries may charge a premium for small orders
Page 12 Transducers 2015
© AMFitzgerald 2015
Not all MEMS will need 200mm wafers Number of Wafers Needed to Produce 10M Units (85% yield) 20000
# Wafers/Year
15000 High‐volume foundry
150 mm 200 mm
10000
5000
0 0
1
2
3 4 5 Die Edge Length, mm
6
7
Pressure sensors, Microphones Accelerometers Page 13 Transducers 2015
Oscillators
Gyros
13 © AMFitzgerald 2015
Choosing your foundry
• The foundry is your partner in a long-term relationship • Switching foundries = starting over ($$$ and time) • Take time and care to make a good decision! “I think this is the beginning of a beautiful friendship.”
Page 14 Transducers 2015
© AMFitzgerald 2015
14
First steps • Sign Non-Disclosure Agreement • Write RFQ “Request for Quote” document – Device drawings, process flow, test requirements – Order quantities, cost targets – Business case
• Send RFQ to multiple foundries • Engineering review with each foundry – Their feedback can save you $ and time!
• Discuss quote with each foundry’s sales team
Process Flow Description
Page 15 Transducers 2015
© AMFitzgerald 2015
The foundry quote 4 Major Phases of Wafer Production
• Quote will be for Feasibility phase only, with estimates for Prototype, Pilot phases – – – – –
NRE: masks, setup, shortloops Wafers Batch size minimum 10 wafers started Priced per wafer or batch No yield guarantees (“best efforts”)
• Typically – multiple Feasibility batches required before moving to Prototyping phase
Feasibility Quality Gate 1
Prototyping Quality Gate 2
Pilot Quality Gate 3
Production
Page 16 Transducers 2015
© AMFitzgerald 2015
Ramp to production timeline (with an existing prototype) FOUNDRY
MEMS COMPANY System development Z loops Week 16
Week 8
Start
Talking/quoting to multiple foundries
Negotiating terms with selected foundry
Week 28
Week 20
Tech transfer
Feasibility wafer batch
Week 32
Week 30 Evaluation and test
Mask revision
N loops, as needed: 12 weeks typ.
Week 32 + (N*12)
Week 40 + (N*12)
Prototype wafer batches
1.5 years, minimum
Pilot qualification: six months minimum
Production
Continuous yield improvement
Wafers from these phases will be useable but may require extra testing Page 17 Transducers 2015
17 © AMFitzgerald 2015
How to get the best prices • Low technical risk keeps NRE low: – Frozen design/process – Repeatable process
• Confirmed order quantities keep wafer prices low: – The more you order, the cheaper it gets
• Show a path to volume – Foundries want to crank out wafers, not do engineering projects Page 18 Transducers 2015
© AMFitzgerald 2015
Making a decision • Evaluate foundry’s process capability and experience with your type of product • Make sure you have compatible: – Business models – Timelines – Expectations – Quality standards – Product volumes – ideally, don’t want to be smallest or largest customer
• Select for best fit, not lowest price Page 19 Transducers 2015
© AMFitzgerald 2015
The foundry will be evaluating you, too!
• Are you a stable company? • Are you well-funded? • Do you have established customers? • Do you know what you are doing? • What are your growth prospects? • Present your business well, don’t make them dig for information Page 20 Transducers 2015
© AMFitzgerald 2015
Closing the deal • Initial deal: – – – –
Price Contract terms Purchase order Downpayment
• Longer term: – Supply agreement – Acceptance terms – End of product life purchase
Page 21 Transducers 2015
© AMFitzgerald 2015
Successful technology transfer and ramp up • Provide to foundry: – Die layout in .GDS – Runsheet, process data
• Dedicate an engineer to be foundry liaison – – – –
Transfer tech info to foundry Monitor wafers in progress Troubleshooting Visit foundry
• Timeline and budget management • To get the best results, be a teammate to your foundry
Page 22 Transducers 2015
© AMFitzgerald 2015
Rookie mistakes • Only quoting one foundry • Expecting to go to production in less than a year • Not presenting a good RFQ or business case • Not discussing your process tolerances with the foundry • Twiddling design/process midstream • Being underfunded • Lack of communication with foundry
Page 23 Transducers 2015
© AMFitzgerald 2015
Fabless challenges • “Golden Wafer” fixation – Just because you made one great wafer in the past, doesn’t mean it can be easily duplicated – You don’t have a device technology dialed until you have identified all the knobs and their settings
• Murphy lives in the fab – Bad, weird stuff will happen, guaranteed – Will you partner with your foundry to solve it, or play the blame game? Page 24 Transducers 2015
© AMFitzgerald 2015
Summary • Selecting a foundry is an extremely important business AND technical decision • Don’t underestimate time and money involved • We can help you! – www.amfitzgerald.com – (650) 347 6367 x101 –
[email protected]
• Join MIG for access to important member resources
Page 25 Transducers 2015
© AMFitzgerald 2015