AI summary by TwinMind
MEMS: The Second Silicon Revolution?
Asianometry
14m 25s
Historical Development & Key Milestones
- Feynman's 1959 lecture "There's Plenty of Room at the Bottom" - offered $1,000 prize for motor fitting in 0.4mm cube
- Prize claimed <1 year later by Caltech engineer using conventional tools
- 1967: Harvey Nathanson at Westinghouse created resonant gate transistor
- ~1mm long, made of gold, electrically controlled gate-electrode distance
- First real MEMS product as frequency filter
- 1979: HP developed silicon micro-machined inkjet nozzle
- Enabled thermal inkjet technology
- Most successful MEMS application up to that point
Inkjet Technology Mechanics
- Rapidly heats ink to ~100°C creating bubbles
- Bubbles push ink through micro-machined nozzle onto paper
- Bubble collapse creates vacuum sucking in more ink
- Nozzles don't move but MEMS enabled tiny size
- Allowed multiple nozzles together for increased printer resolution
1990s Sensor Revolution
Automotive Airbag Sensors
- Original sensors: mechanical balls in tube - large, heavy, complex, expensive ($100s)
- 1991: Analog Devices created first commercial acceleration MEMS sensor
- <1 square centimeter size
- 60 capacitive sense fingers, 60 microns deep
- Fingers attached to mass that moves during acceleration
- Touching fingers change capacitive value sensed by chip
- ADXL50 chip (50G acceleration capability)
- Complete monolithic system: smaller, more reliable, $5 vs $20-100
- >60 million MEMS airbag sensors made/sold
- Cars now have dozens of MEMS sensors: anti-lock brakes, active suspension, navigation, rollover detection
Medical Pressure Sensors
- Monitor blood pressure via intravenous line
- Early external sensors: >$600, required sterilization/recalibration per use
- MEMS version: $10, majority cost from packaging, disposable
- Mechanics: silicon thin membrane diaphragm with 4 stress-sensitive polysilicon/piezo resistors at edges
- Diaphragm suspended over vacuum cavity for absolute pressure sensing
- Pressure changes silicon band structure, changes crystal resistivity
- Medical MEMS sensor industry: >$6 billion, blood pressure sensors largest segment
Consumer Electronics Applications
- Nintendo Switch: ultra-low power inertial sensors with gyroscope/accelerometer (STMicroelectronics SH-627)
- Visual displays, optical switches using micro mirrors
- Micro mirrors redirect light between optical fibers
Manufacturing Technologies & Approaches
- Primary approach: high-volume IC techniques (photolithography, etching)
- Add/remove layers on 2D substrate to produce 3D shapes
- No one-size-fits-all methodology
- Alternative approaches:
- Electrochemical micro machining (ECM) - originally for PC boards
- Laser beam machining
- Electrode discharge machining
- LIGA (unpronounceable acronym)
Design Challenges vs Traditional ICs
Fundamental Differences
- Cannot simply miniaturize and expect same performance
- Friction forces much stronger than inertia at micro scale
- Tiny flow spaces prone to fluid blockages
- Electro-osmotic effects enable electric field pumping at micrometer scale
Design Philosophy Differences
- VLSI uses modular approach with generic modules/standardized cell libraries
- Libraries designed independently from product (e.g., TSMC nodes)
- Mechanical design lacks standardization except bolts, fasteners, pipe fittings, paint
- Mechanical components must accommodate weight/space/energy restrictions for specific products
- VLSI objects process signals (logic only)
- Mechanical objects: contain liquids, rotate, slide, support loads simultaneously
- Makes MEMS design extremely custom
Design Tools
- CAD/design automation software: Coventreware, Intellicad
- Help designers understand microstructure design and electromechanical behaviors
Packaging & Cost Structure Issues
Cost Breakdown
- Packaging: 75-80% of final MEMS cost (reversal from IC situation)
- HVAC pressure sensor: $20-30 total price
- Raw die from foundry: <$0.50 (unusable without packaging)
Fundamental Packaging Challenges
- ICs are self-contained systems - packaging provides protection/interface
- MEMS must intimately interact with outside world
- Cannot encase pressure sensor in hard plastic - wouldn't sense pressure
- Each MEMS type needs different packaging:
- Pressure sensor ≠ micro mirror array
- Microfluidic ≠ electrostatic actuator
- Adds substantial cost and engineering overhead
Economic Challenges & Market Dynamics
Analog Devices Case Study
- Started MEMS research 1980s targeting airbag sensors
- Built in-house manufacturing facility
- Took 9 years for facility to become profitable
- Marketed significant price advantage over competition
- First automotive deal with Siemens: automatic price declines over 6 years
- $11/unit declining to $6/unit
- Expected efficiency gains similar to IC industry
Yield & Pricing Issues
- Yield is biggest cost factor - 10% yield means 90% unusable products
- Doubling yield halves cost
- Difficult to achieve, leads to overly optimistic projections
- Market prices don't reflect actual investment in development/production/packaging/testing
- Relentless price erosion: 3-5% per quarter
- Erodes margins, disincentivizes investment in new applications
Missed Opportunities
- Micropropulsion for cube satellites
- MEMS-based LIDAR mirrors for autonomous cars
- IoT sensors
Industry Structure & Future Outlook
- Called "second silicon revolution" - microscopic moving parts, smaller/cheaper/better performance
- Victim of own success - creativity/diversity makes standardization harder
- No industry coordination like SEMATECH or DARPA
- ~50 players: designers, foundries, vendors, end-users
- Need closer collaboration for standards and solutions to under-invested issues
- Required for realizing full technical/commercial potential
