AI summary by TwinMind
The iPhone Forever Changed the RF Filter
Asianometry
22m 21s
RF Filter Fundamentals
- RF filters receive input signals, create output based on frequency
- Pass band: desired frequency range passes through with minimal insertion loss
- Stop band: unwanted frequencies rejected (ideally zero output)
- Filter response resembles "woman's skirt" or mesa shape
- Quality factor (Q): ratio of center frequency to bandwidth
- Higher Q = better performance
- Lower insertion loss = better performance
- RF filters operate 100 MHz to 10 GHz (above 10 GHz = microwave filters)
Surface Acoustic Wave (SAW) Filters History
- Concepts date to 1880s, Lord Rayleigh predicted/created Rayleigh waves formula (1885)
- Surface acoustic waves travel across solid surfaces (seismic waves similar)
- WWII heightened SAW research for radar signal-to-noise improvement
- 1965: UC Berkeley Professor Richard White + graduate student F.W. Volmer demonstrated effect
- Used interdigitated transducers (IDTs) - comb-shaped metallic electrodes
- Two IDTs placed on crystalline quartz bar (piezoelectric substrate)
- Input IDT converts radio signal to mechanical energy → surface acoustic wave
- Output IDT converts wave back to electrical signal via piezoelectric effect
- Some devices add reflectors creating surface acoustic wave resonators
SAW Filter Manufacturing
- MEMS devices (microelectromechanical systems)
- 1980s: gaps between IDT electrodes as small as 300 nm
- Process: deposition of aluminum alloy thin film on piezoelectric substrate
- Substrate: quartz crystal or lithium niobate for higher frequencies
- Lithography patterns features, then etching
- Early SAW filters: single layer manufacturing
- Advanced SAW filters: more complex but still minimal layers
- Packaging challenge: often sealed in vacuums
Early SAW Applications
- Military first to research for communication systems
- Used on Voyager missions (late 1970s) - worked nearly 50 years
- 1975: TV companies (Phillips, Pye, Siemens) adopted for low-cost replacements
- Replaced coil/capacitor-based filters
- Better tuning, reduced static interference
- 30-40 million units produced - first big MEMS products
- 1973: Motorola DynaTAC used ceramic filters (larger than SAW)
- 1992: Fujitsu fabricated first SAW resonator RF filters for mobiles
- Small, cheap, easy manufacture
- 1975: ~$2 each
- 1990s: $1.40-$1.60 per filter (50 million unit volumes)
SAW Filter Limitations
- Challenging above 2 GHz frequencies
- Higher frequencies require smaller electrode gaps (too small to produce easily)
- GSM evolution: started single band (2 filters) → multiple bands → filter interference
- Need for new solution by 3G era
Bulk Acoustic Wave (BAW) Filters Development
- 1998: German semiconductor maker Infineon began BAW development
- BAW concepts date to 1980
- Energy travels through entire bulk (not just surface)
- Waves travel faster, farther without deterioration
- Two subtypes:
- Film Bulk Acoustic Resonator (FBAR)
- Solidly Mounted Resonator (SMR)
- Core structure: capacitor with top/bottom electrodes sandwiching piezoelectric material
- Center frequency determined by piezoelectric layer thickness
- 10 MHz crystal: ~170 micrometers thick with demanding tolerances
FBAR vs SMR Differences
- FBAR: suspended sandwich in midair with literal air gap
- Uses sacrificial layer technique, etched away with acid/gas
- Variations: thin membrane, springs, vias at edges
- SMR: solidly grounded with acoustic reflectors (Bragg reflectors)
- Can use traditional VLSI technology
- Generally less performant than FBAR at high frequencies
BAW Filter Advantages/Disadvantages
- Advantages over SAW:
- Better for higher frequencies
- Capacitor structure prevents electrical cross-talk
- Less sensitive to surface contamination
- Disadvantages:
- Much harder to manufacture
- More complicated structures
- Simple BAW: 1 mask layer (poor performance)
- Good FBAR: 9-13+ mask layers vs SAW's 1-3 layers
- Materials: aluminum nitride or zinc oxide piezoelectric layers
Industry Evolution
- HP Labs began FBAR research (1993)
- HP spun off Agilent, first shipped commercial duplexers with FBAR filters
- 2006: KKR/Silver Lake bought Agilent semiconductor → renamed Avago
- 2008: Avago bought Infineon FBAR group ($20 million)
- 2015: Avago bought Broadcom
- Current: Avago now operates as Broadcom
iPhone Revolution Impact
- Original iPhone (2007): 2G radio, quadband GSM/EDGE, used SAW filters
- iPhone 3G (2008): 3G GSM, continued SAW filters (possibly Murata)
- iPhone 4/4S: SAW filters from Skyworks and TriQuint (now Qorvo)
- iPhone 5 (2011): First LTE iPhone - "everything changed"
- LTE: 40 frequency bands globally (600 MHz to 3,600 MHz)
- Required filters for all bands
- TriQuint duplexers for UMTS, Avago FBAR duplexer for LTE
- Apple's "world phone" strategy for global compatibility
Market Explosion
- Original iPhone: quadband radio
- iPhone 7: 23-24 GSM/CDMA/LTE bands
- Samsung Galaxy: 16 bands
- Plus Wi-Fi, GPS, Bluetooth, NFC
- 2004: BAW RF filter market <$100 million
- 2016: High-band RF filter industry >$1.6 billion
- Market pressures: thinner, lighter, more performant devices
- Integration of filters, switches, power amplifiers into single RF units
Current Market Structure
- BAW filter industry: 95% controlled by Qorvo and Broadcom
- Broadcom fabs filters in Fort Collins, Colorado (~1,100 employees)
- SAW market: multiple players including Murata, Skyworks, Qorvo, Taiyo Yuden
- Dozens of smaller vendors
5G Challenges
- 5G demands higher frequencies and larger frequency slices
- Requires even thinner electrode/piezoelectric sandwiches
- Material changes needed
- Aluminum nitride standard being supplemented with scandium doping
- Improves RF performance
- Creates manufacturing challenges
- Companies experimenting with optimal scandium ratios
Future Outlook
- 2019: Broadcom considered selling RF business for $10 billion (didn't sell)
- Industry belief: 5G won't match 4G/LTE boom
- iPhone 15 Pro: supports ~40 LTE/5G bands
- Challenges: manufacturability, declining returns
- Continued demands: new frequencies, integration, smaller sizes, better heat/power
- Growth days of early 2010s likely over
