Product Opportunities and Challenges. Commercial RF-MEMS: wi Spry. Arthur S. Morris, III CTO, VP Eng.

Transcription

1 Commercial RF-MEMS: Product Opportunities and Challenges Arthur S. Morris, III CTO, VP Eng.

2 Introduction Who is wispry? Spun out from Coventor at end of 2002 Developing RF-MEMS for services customers since 1999 High performance low-cost RF components Broad customer evaluations underway Agenda MEMS Technology Update Promising Application Opportunities Key Challenges and Solutions Summary

3 Micromachining Transmission Lines Inductors Maintain 50Ω at small dimensions With losses < 1dB/cm at 10 GHz L > 14 nh Q up to 100 at 2 GHz Membrane Supported Lines U Michigan Micromachined Waveguides KAIST JPL, CIT, IEEE U Michigan

4 Micromechanical Resonators & Filters Thin-film Bulk Flexural Mode RF up to 7.5 GHz Q > 1000 Agilent shipping FBAR-based SAW-replacement filters. IF above 90 MHz & Q > 7000 Future >300 MHz & Q > Discera sampling oscillators stabilized with these resonators.

5 MEMS Varactors and Switches Capacitance > 8 pf Ratios > 10:1 Q up to 200 at 2 GHz Low Loss < 0.1dB High Isolation > 50 db Ultra-linear Raytheon OFF (transmitting) ON (reflecting) BSAC and LLNL UC Boulder Magfusion RF out Bias Electrode HRL University of Hawaii Anchor Contact Dimple HRL RF in

6 RF-MEMS Switches High Isolation Physical gap Low Insertion Loss All metal pathway Low Power Electrostatic Actuation (typ.) Ultra-linear Metal-Metal contacts Property MESFET PIN Diode PHEMT MEMS Series Resistance 3 Ω 5 Ω 1 Ω 3 Ω 1 Ω 3 Ω <1 Ω Loss at 1 GHz 0.5 db 1.0 db 0.5 db 0.8 db 0.3 db 0.8 db 0.1 db Isolation at 1 GHz 15 db 30dB 30 db 50 db 20 db 35dB >50 db IP dbm dbm dbm >70 dbm 1 db comp dbm dbm dbm >37 dbm 1 Size 1-5 mm mm mm 2 <0.1 mm 2 Switching Speed ~ns ~ms ~ns 1-100µs Control Voltage 8 V 3-5V 2.3 to 5 V 2-100V Control Current < 10 µa 1 ma < 30 µa <<1 µa MEMS Provide Low Loss, High Isolation, High Linearity and Small Size Simultaneously! 1 No compression observed NB: Typical values

7 Opportunities Abound in RF Front Ends Reconfigurable Filters Tracking Filters Low-loss Filters Filter bank switching T/R switching Amp stage bypass Amp dynamic matching Amp band switching Antenna tuning Antenna reconfiguration Antenna phase shifters Tunable oscillators Mixers LNA LNA LNA LNA LNA LNA MEMS Opportunities VGA VGA div 2 I Q I div 2 Q MHz DET LPF LPF LPF LPF MHz ADC ADC ADC ADC RXIDATA_CDMA (serial or parallel) RXQDATA_CDMA (serial or parallel) RXIDATA_FM (serial or parallel) RXQDATA_FM (serial or parallel) I div 2 Q I div 2 Q DAC DAC DAC DAC (IC VCO, ext. tank) (IC VCO, ext. tank) (VCTCXO) (IC VCO, ext. tank) (IC VCO, ext. tank) TXIDATA (serial or parallel) PLL TXQDATA (serial or parallel) TXIDATA (serial or parallel) PLL PLL PLL TXQDATA (serial or parallel)

8 Toward Re-configurable RF Front-Ends Optimal Integration of RF Architectures VCO Dual-band VCO Multi-band Low-Noise VCO Tunable Transmit Filter Transmit Filters Tri-band PA Multi-band PA Filtered Multi-band PA (may be integrated with transceiver at low power) Filtered Multi-band PA wispry s Initial Products Multi-band Transceiver VCO Dual-band VCO SAW Transmit Filters Tri-band PA Base band Base band Transceiver Switchplexer Fixed Omni Antenna Transceiver Tunable Directional Transceiver w/ Multi-band VCO Multi-band LNA & Diplexer Monolithic Switchplexer Antenna Switchplex er Fixed Omni Antenna SAW Receive Filters Tunable Receive Filter SAW Receive Filtered Multi-band Filters Multi-band LNA LNA Tri-band LNA Tri-band LNA Benefits Lower Overall Cost (Savings of 50-75% Possible) Smaller Footprint Improved RF Performance Faster Time to Market

9 s Initial Focus Switches Band/Mode/Path/Chain Selection Beyond SP6T in < 2 mm 2 2-bit transfer Matrices High-current options Monolithic Switchplexer Band Diplexer Transmit / Receive Switching Phase Shifter Smart Antennas Test and measurement

10 Typical Switch Die S-parameters Reflection, Loss (db) Isolation, Return Loss (db) 2.4GHz Input Reflection Coefficient Loss Return Loss Isolation Open Ref lection (db) Loss (db) Isolation (db) Retur n Loss (db) freq (66.08MHz to 26.50GHz) Reflection Frequency (GHz) Non-deembedded results on low resistivity silicon! S(1,1)

11 Packaged S-parameters Customer Tests

12 Large-Signal Switch Measurements Negligible harmonics to 4 watts Fo 3Fo Clean ACLR at 1 watt Harmonic Level (-dbc) Output Fo (dbm) Negligible RF actuation to 4 watts Isolation (db) Input Fo (dbm) Isolation

13 db(lpf_5r8g_e1_square_2945_upper..s(1,1)) db(lowpass_filter_2l3c_6_ghz..s(1,1)) First-Pass Filter Performance Direct Comparison: Simulation (Qind = 15) vs Measured LPF_2r4G_E1 (square inductors, solid-oxide beneath (not-sacrificed)) m3 freq=2.450ghz 2.4 GHz LPF db(lowpass_filter_2l3c_2r5_ghz..s(2,1))= m4 freq=2.445ghz db(lpf_2r4g_e1_square_2854_upper..s(2,1))= m3 m4 5 0 db(lpf_2r4g_e1_square_2854_upper..s(1,1)) db(lowpass_filter_2l3c_2r5_ghz..s(1,1)) Direct Comparison: Simulation (Qind = 15) vs Measured LPF_5r8G_E1 db(lpf_2r4g_e1_square_2854_upper..s(2,1)) db(lowpass_filter_2l3c_2r5_ghz..s(2,1)) (square inductors, solid-oxide -35 beneath (not-sacrificed)) m3 5.8 freq=5.800ghz -40 GHz LPF db(lowpass_filter_2l3c_6_ghz..s(2,1))= m4 freq=5.813ghz db(lpf_5r8g_e1_square_2945_upper..s(2,1))= Lowpass filters from same wafers as switches Sub-component of switchplexer Modeled design and measured data agree Insertion loss high low resistivity substrate Q =~ 2.4 GHz Q =~ 5.8 GHz freq, GHz m3 m db(lpf_5r8g_e1_square_2945_upper..s(2,1)) db(lowpass_filter_2l3c_6_ghz..s(2,1)) Expect Q > 50 with improved design Meet specs in next run freq, GHz (Red and blue = simulation; pink and lt.blue = measurement)

14 Phase Shifter Smart Antenna Applications Single Chip, Chip-Scale Packaged ~ 2 mm 2 Through loss < 0.2 db per bit Integrated Low Loss Delay Lines Transmission Lines at High Frequencies Compact LC Delay Line at Low Frequencies (800MHz 6 GHz) 5.8 GHz 2.4 GHz

15 Key Challenges and Solutions Why are RF-MEMS not already in the mainstream? Cost MEMS dominate accelerometer and projection display markets Hermetic sealing and packaging Size of overall solution Integration barriers Insufficient foundry volumes Reliability Design/process specific issues Reliability has been unproven Control High voltages or currents have been required Switching time slower than solid-state

16 Wafer-Level Encapsulation Lower Cost Cavity packaging expensive Higher yield Lower labor costs Easier Integration MCM Flip-chip Monolithic Seal ring Longer Life Sealed in fab clean room Controlled atmosphere Higher Performance Controlled Impedance Chip-on-board, etc. Reduced contamination Buried low-loss interconnect

17 Packaging Approaches Protos and PCB Mounting (DRS1) 2 x 2 x 1.2mm package 400um Pitch; 200um Balls 25 Balls on 5 x 5 Grid Module Mounting (ERS1) 1.5 x 1.5 x 0.5mm WLCSP 200 um Pitch; 75um studs In Development 36 pins on 6x6 grid Solder Ball on Plated Via Wafer-Level Bond Next-gen Module Mounting (LRS1) 1.0 x 1.0 x 0.3 mm WLCSP In Research 150 um Pitch; 50um studs 32 pins; Quad Configuration Thin-Film Cap

18 Integration within a single process Extensive Reuse of Proven Components Basic Component Set Switches Transmission Lines Multi-layer Capacitors Inductors Transformers Variable Capacitors Variable Inductors Post-processing of Actives Charge Pump Circuit RF-CMOS + + = Unique RF-MEMS Process Low k & tanδ glass Thick Copper and Gold No Polysilicon Low Thermal Budget Wafer-Level (Chip-Scale) Packaging Stand-Alone or Post-Process Any IC Wafer High performance on low or high resistivity silicon Available in Multiple Foundries Complete Solution

19 Thorough Design for Reliability Device Operation Contact and Release Forces Contact Materials Electric Fields and Charging Mechanical Materials Mech. Stress Concentrations Defect Analysis Current Handling Thermal Stability Vibration Standoff Position Lb = 0 Lb = 20 Standoff Touchdown Actuation Voltage (V) Extremes Overvoltaging ESD RF Pull-in Pulsed Current Shock Packaging Sealing Atmosphere Gettering Stresses Release Force Parasitic Actuation One Bump Contact Force (un)

20 Switch Life Cycling Operate to billions of cycles without failure unpackaged Contact stable with <1 mω increase per million cycles Numerous switches tested using automated equipment Reliability testing of fully packaged parts is continuing Acceleration method search underway Contact Pair Resistance (Ω) Log 10 Cumulative Cycles

21 Charge Pump 2.7V supply Regulated +/- 15V Control Includes Voltage converter Switch drivers CMOS/TTL Interfaces 35uA available per switch for worst case of continuous switching operation Shutdown During Stand-By Power Up/Down <100usec Discrete version built and tested Beginning CMOS Design CTL1 (supply Vdd for pump 1?) CTLn (supply Vdd for pump n?) OR Vdd (or wire- OR'ed inputs) Oscillator / Buffer GND Clock Generator 1 2 m Charge Pump 1 Charge Pump n OUT1+ (OUT1-/ GND) OUTn+ (OUTn-/ GND)

22 Transient Response Closing time: <100 µs to stable resistance <50µs delay, <25µs 10%-90% RF transition Opening time: < 10 µs 6.0E E E E-01 Open BBL2 BBL2 6.0E-01 Actuation 60V 5.0E-01 Closing 70V 4.0E E-01 Open Actuation 60V Opening 70V V (V) V (V) 2.0E E E-01 Closed 0.0E E E E E E E E-04 Time (s) 1.0E E+00 Closed 5.00E E E E E-03 Time (s) Fast enough for many applications, even some T/R Air damping limited, easily lowered by an order of magnitude

23 Summary Low cost overall solution is key Wafer-level chip-scale packaging True relay -> No off-chip components required for DC-RF Isolation Proprietary Unique IC Compatible Process Proven; Silicon Since Late 2000; Ported and Installed in 3 Fabs All Products on Same Process a la Mixed-Signal ICs; Aggregation of Volume High reliability achieved DFM Methodology Integrated with RF IC Flow (Cadence) Proven to > 10 9 cycles Acceleration mechanisms under study Control issues resolved Actuation Voltages are <±10V with Path to Lower Voltages Less Demanding Charge Pump Design -> Easier direct integration Speed sufficient for most commercial applications MEMS are ready for prime-time in RF applications Proliferation of wireless standards increases need Technology is maturing

24 Questions?