SMART DUST B. Boser, D. Culler, J. Kahn, K. Pister Berkeley Sensor & Actuator Center Electrical Engineering & Computer Sciences UC Berkeley SMART DUST Outline • History • Technology Ramblings SMART DUST Motivation • Exponential decrease in size, power, cost • Digital computation • Analog/RF communication • Sensors battery Goals • Understand fundamental limits • Build working systems SMART DUST Moore’s Law, take 2 • Nanochips on a dime (Prof. Steve Smith, EECS) SMART DUST DoD Workshops • RAND 1992 • “Future Technology-Driven Revolutions in Military Conflict” • “Smart Chaff”, “Floating Finks” • Bruno Augenstein, Seldon Crary, Noel Macdonald, Randy Steeb, … • Santa Fe, 1995 • Xan Alexander, Ken Gabriel; Roger Howe, George Whitesides, … • ISAT 1995, 1996, 1997, 1998, 1999, 2000 •… SMART DUST University Programs (old slide) • UCLA • Bill Kaiser (LWIM, WINS) • Greg Pottie (AWAIRS) • U. Michigan • Ken Wise • USC • Deborah Estrin • UCB • K. Pister (Smart Dust) •… SMART DUST • Ken Wise, U. Michigan SMART DUST http://www.eecs.umich.edu/~wise/Research/Overview/wise_research.pdf Bill Kaiser, UCLA • http://www.janet.ucla.edu/WINS SMART DUST August ’01 Goal SMART DUST COTS Dust - RF Motes • Simple computer • Cordless phone radio • Up to 2 year battery life N W E S 2 Axis Magnetic Sensor 2 Axis Accelerometer Light Intensity Sensor Humidity Sensor Pressure Sensor Temperature Sensor SMART DUST COTS Dust GOALS: • Create a network of sensors • Explore system design issues SMART DUST COTS Dust RESULTS: • TinyOS – David Culler, UCB • Manufactured by Crossbow ~ $150 • 100+ users, 40+ locations • Military and civilian applications SMART DUST 800 node demo at Intel Developers Forum 4 sensors $70,000 / 1000 Concept to demo in 30 days! SMART DUST Structural performance due to multi-directional ground motions (Glaser & CalTech) . Mote Layout 1 3 1 54 6` 1 8 1 1 1 5 29 Mote infrastructure Comparison of Results Wiring for traditional structural instrumentation + truckload of equipment SMART DUST Cory Energy Monitoring/Mgmt System • 50 nodes on 4th floor • 5 level ad hoc net • 30 sec sampling • 250K samples to database over 6 weeks SMART DUST 29 Palms Sensorweb Experiment • Goals • Deploy a sensor network onto a road from an unmanned aerial vehicle (UAV) • Detect and track vehicles passing through the network • Transfer vehicle track information from the ground network to the UAV • Transfer vehicle track information from the UAV to an observer at the base camp. SMART DUST Flight Data SMART DUST Dragon Wagon HMMWV HMMWV From UAV Dragon Wagon From UAV SMART DUST Last 2 of 6 motes are dropped from UAV • 8 packaged motes loaded on plane Last 2 of six being dropped SMART DUST Detection algorithm Each vehicle V(v,t) has two parameters: 1) Speed (v) 2) Time at beginning of network (t) The n-node network is described by an n-entry pattern vector p: The jth entry is the time we expect that node j will see V(1,0) Times when nodes detect V are collected in the t vector p1 p 2 p3 p 4 1 t1 1 1 / v t 2 1 t t3 t 1 4 Ax b Linear least-squares guess at v and t SMART DUST Room to spare! SMART DUST RF Sensitivity • Pn = kBT Df Nf • Sensitivity = Pn + SNRmin • e.g. GSM (European cell phone standard), 115kbps k BT 200kHz ~8x SNR S = -174dBm + 53 dB + 9 dB + 10 dB = -102 dBm RX power = ~200mW TX power = ~4W 50 uJ/bit SMART DUST RF Path Loss • Isotropic radiator, l/4 dipole • Pr=Pt / (4p (d/l)n) • Free space n=2 • Ground level n=2—7, average 4 SMART DUST N=4 From Mobile Cellular Telecommunications, W.C.Y. Lee Pt = 10-50W -102dBm SMART DUST Path Loss • Like to choose longer wavelength • Loss ~(l/d)n • 916MHz, 30m, 92dB power loss • need –92dBm receiver for 1mW xmitter • power! • Penetration of structures, foliage, … • But… • Antenna efficiency • Size – l/4 @ 1GHz = 7.5cm SMART DUST Output Power Efficiency • RF • Slope Efficiency • Linear mod. ~10% • GMSK ~50% • Poverhead = 1-100mW Pout True Efficiency Slope Efficiency • Optical • Slope Efficiency • lasers ~25% • LEDs ~80% • Poverhead = 1uW-100mW Poverhead Pin SMART DUST Limits to RF Communication Cassini • 8 GHz (3.5cm) • 20 W • 1.5x109 km • 115 kbps • -130dbm Rx • 10-21 J/bit • kT=4x 10-21 J @300K • ~5000 3.5cm photons/bit Canberra • 4m, 70m antennas SMART DUST Video Semaphore Decoding Diverged beam @ 5.2 km In shadow in evening sun SMART DUST ~8mm3 laser scanner Two 4-bit mechanical DACs control mirror scan angles. ~6 degrees azimuth, 3 elevation 1Mbps SMART DUST Application to Microassembly • • Pattern complementary hydrophobic shapes onto parts and substrates using SAMs. • no shape constraints on parts • no bulk micromachining of substrate • submicron, orientational alignment Uthara Srinivasan, Ph.D. thesis, UC Berkeley Chem.Eng., May 2001 Courtesy: Roger Howe, UCB SMART DUST Mirrors in Solution Courtesy: Roger Howe, UCB SMART DUST Mirrors on Microactuators assembled mirror Courtesy: Roger Howe, UCB SMART DUST CMOS Imaging Detector Photosensor CRC Check SIPO Shift Local Bus Driver Register Signal Processing A/D Conversion Off Chip Bus Driver Pixel Array SMART DUST Power and Energy • Sources • Solar cells ~0.1mW/mm2, ~1J/day/mm2 • Combustion/Thermopiles • Storage • Batteries ~1 J/mm3 • Capacitors ~0.01 J/mm3 • Usage • Digital computation: nJ/instruction 10 pJ • Analog circuitry: nJ/sample 27 pJ/sample • Communication: nJ/bit 11 pJ RX, 2pJ TX SMART DUST Smart Dust - Processes (CMOS) 330µm TX Drivers Power input ambient light sensor Photodiode Sensor input ADC 70kS/s, 1.8uW 0-100kbps CCR or diode Power Oscillator 13 state FSM controller Optical Receiver 1 Mbps, 11uW 1mm What’s working – Oscillator, FSM, ADC, photosensor, TX drivers What’s kind of working – Optical receiver (stability problems lead to occasional false packets) SMART DUST Power, sensor, motor fab Isolation trenches are etched through an SOI wafer and backfilled with nitride and undoped polysilicon. SMART DUST Power, sensor, motor fab Solar cells and circuits are created by ion implantation, drive-in, oxidation, contact etching, aluminum sputtering and etching. SMART DUST Power, sensor, motor fab Actuators are deep reactive ion etched through device layer. SMART DUST Power, sensor, motor fab Optional backside etch (would actually precede front side etch) SMART DUST Solar Cell Results 0.5 to 100 V demonstrated 10-14% efficiency Solar Array Performance 0.2 Current (uA) 0.1 0.0 0 5 10 15 20 25 30 -0.1 -0.2 -0.3 -0.4 Voltage (V) SMART DUST Power from MEMS Combustion Nozzle (w/ igniter) Thermopiles SMART DUST Closing in on 1mm3 2.8mm 2.1mm Solar Cells CCR Accelerometer CMOS IC SMART DUST Smart Dust - Integration Solar Cell Array CCR XL CMOS IC SENSORS ADC PHOTO 8-bits 1V 1-2V FSM RECEIVER 375 kbps 16 mm3 total circumscribed volume TRANSMITTER 175 bps ~4.8 mm3 total displaced volume 1V 1V SOLAR POWER 3-8V 2V OPTICAL IN OPTICAL OUT SMART DUST 175 bps from 10 mm3 CCR Drive Voltage Sample from XL pad (connected to Vdd) Echo of Downlink data Sample from photosensor Detected Transmission SMART DUST Mote with Micro-battery from Lee & Lin, UCB SMART DUST Optical Communication Path loss 0-25% 25% Loss = (Antenna Gain) Areceiver / (4p d2) Antenna Gain = 4p / q½2 SMART DUST Theoretical Performance 5km Photosensor CRC Check SIPO Shift Local Bus Driver Register Signal Processing A/D Conversion Off Chip Bus Driver Ptotal = 50mW Pt = 5mW q½ = 1mrad BR = 5 Mbps 10nJ/bit Pixel Array Areceiver = 1cm2 Pr = 10nW (-50dBm) Ptotal = 50uW SNR = 15 dB ~10,000 photons/bit SMART DUST Theoretical Performance 5m Ptotal = 100uW Pt = 10uW q½ = 1mrad BR = 5 Mbps Areceiver = 0.1mm2 Pr = 10nW (-50dBm) Ptotal = 50uW SNR = 15 dB 20pJ/bit! SMART DUST RF mote • CMOS ASIC • 8 bit microcontroller • Custom interface circuits • External components Temp ~$1 uP SRAM Amp ADC Radio ~2 mm^2 ASIC battery antenna inductor crystal SMART DUST Radio basics • Tuneable frequency, 900MHz +/-100 MHz • Programmable power output • -10 – 0 dBm out, 1 – 10 mW in • 100 kbps? 13 bit freq. reg. uP Tuneable cap. Oscillator core 8 bit power reg. Tuneable power SMART DUST Radio basics • Tuneable frequency, 900MHz +/-100 MHz • Programmable sensitivity • -100 – -90 dBm, 0.1 – 10 mW in • 100 kbps? • Many interface options • Direct memory • Low power vigilance? DMA pointer uP Oscillator core SRAM SMART DUST Crystal-free radio? • ~20% variation in frequency reference in • • • CMOS I measure your frequency output in my coordinate system, and vice versa Theory of coupled oscillators Digital feedback between nodes SMART DUST Wakeup synchronization • Watch crystals • 32kHz, 30nW • 10-100 ppm drift • 1-10 ms/min • 1-10 sec/day • 5-50 min/year • Temperature is primary source of drift • Compensate to sub-ppm – 100ppb? SMART DUST RF Mote Summary • Available 2003 • Radio • 900 MHz • 10+ m range • 10 nJ/bit (0.3mA, 100kbps) • 8 bit Atmel-ish uP • 10pJ/inst (0.03mA) • 10 bit ADC • 100kS/s, 30nJ/sample (0.01mA) • Batteries • Lithium coin cell ~ 220mAh • AA batteries 1000mAh SMART DUST Abstracting the Hardware • Goal: • Provide realistic energy (and time) metrics to drive algorithm development • Allow software/algorithms to drive hardware design. Distributed localization Centralized localization Diffusion routing … Routing tables Abstract representation of hardware Rene mote Mica mote Laptops & Wavelan SMART DUST Abstracting the Hardware • Too simple: • “computation” = x pJ • Comm = y nJ/bit*m^4 • Sensing = z pJ/sample • Too complex: • 16 bit add register to non-cached main memory = x pJ, • … SMART DUST Abstracting the Hardware • Need a representation(s) of • Energy cost • Latency • Probabilistic? SMART DUST Example: maximize sensor net lifetime • Given: • Costs of sensing, computation, communication • Fixed sensor locations • Connectivity matrix • One or more base stations • Find: • Energy-optimal routing to get data back from each node (define it first!) • Everyone on all the time • Duty cycling SMART DUST Example: minimal coverage • Given: • Costs of sensing, computation, communication • Sensor range, communication range • Mote weight dominated by battery • Find: • Minimal dispersion of motes (in kg/km2 !) st. events x,y,z can be sensed for time t SMART DUST Example: minimal coverage • Workstation? SMART DUST Example: minimal coverage • Smart dust? SMART DUST Example: minimal coverage • Some of both? SMART DUST Mobility SMART DUST Other topics • Simulation of big networks • Data fusion/compression • Information theory • Shannon for sensor networks • What is “capacity”? • Collaborative signal processing • Definition • Existence? SMART DUST Summary • Cubic-inch RF motes working in applications • 10 mm3 optical motes demonstrated • 10 mm3 RF motes coming • Peer-to-peer networking • Most communication is relay • Energy cost to communicate 1 bit is at least 1000x greater than an 8 bit instruction SMART DUST Conclusion 1 3 mm or bust!!! SMART DUST