The Mote Revolution: Low Power Wireless Sensor Network Devices Joseph Polastre, Robert Szewczyk, Cory Sharp, and David Culler {polastre, szewczyk, cssharp, culler} @ cs.berkeley.edu Computer Science Department, University of California, Berkeley A family of low power wireless sensor network devices have been built to enable research and deployments. The devices have featured commercial off the shelf (COTS) components integrated together on a platform commonly referred to as a "mote" [2,4]. Motes have been used to evaluate wireless sensor network algorithms as well as for environmental monitoring and object tracking deployments. Miniature wireless devices are ideal for high density long term deployments in areas otherwise unsuitable for wired connections or passive devices. We describe the design and implementation of a next generation sensor mote that utilizes emerging hardware, miniature sensors, and new standards to achieve high data rate, extremely low power operation for monitoring applications [6]. Fine grained power management is essential for wireless sensor network applications that run for months (such as an organism's breeding season) or years (in HVAC or structural monitoring). Low power operation is achieved not only through selection of efficient hardware, but also through duty cycling that hardware; care must be taken to minimize common operations. From experience building a family of wireless sensor nodes, or "motes" (see Table 1 attached), we have observed the primary operations in low duty cycle operation are sleep, wakeup, and run (in that order). One cycle of sleep, wakeup, and run is typically the cost of acquiring a single set of sensor samples. For the majority of the time the node is sleeping. While asleep, the microcontroller must maintain its state (typically in RAM) while consuming little power and shutting down or disconnecting all peripherals including the radio. For Telos, our newest and lowest power mote to date, we chose the Texas Instruments MSP430 microcontroller (specifically the F1611 although our device is backwards compatible with the F149) [7]. The MSP430 consumes only 2 microwatts in sleep mode while maintaining RAM. Other microcontroller features include direct memory access, supply voltage supervisor, and hardware implementations of SPI, USART, and I2C. In low power applications, it is important to run all auxiliary hardware components (ADC, DMA, bus operations) from a low speed (often 32kHz) external clock so that the microcontroller oscillator core can be shut down for additional power savings. During the wakeup phase, the mote must start the core of the microcontroller for processing. It may also start the radio and sensors for sampling. The MSP430 can wake up in under 6 microseconds. Short deterministic wakeups are important for two types of low power communication--low power listening and scheduled communication. Low power listening periodically samples the radio channel for activity. If no activity is found, the node returns to sleep. Low power listening does not require time synchronization, instead the duty cycle is dependent on the network traffic and the startup time of the microcontroller. Scheduled communication performs synchronization with the network and wakes up at specified intervals for communication. Minimizing the active time with a scheduled communication scheme requires minimizing the non-deterministic portion of the startup time. The active or running portion of the mote's duty cycle is typically an extremely small portion of the overall mote lifetime. Accordingly, we first focused on selecting components with low sleep profiles and fast wakeups. When the node is active, it is important to be able to control the power to each of the external peripherals. Keeping the active time small requires a low power yet fast microcontroller. The MSP430 is a 16-bit microcontroller running at 4MHz with only 0.5 milliwatts active current consumption. We chose to couple the MSP430 with the CC2420 [1] IEEE 802.15.4 [5] compliant radio to enable standardized network communication between different hardware devices. Previous to IEEE 802.15.4, motes used radios with custom protocol implementations. Now with IEEE 802.15.4, sensor networks may grow to very large networks through interoperating devices with a standardized (and FCC certifiable) radio and MAC protocol layer. The CC2420 is a low power transceiver (17mA receive/19mA transmit) at 250kbps using O-QPSK. The startup time of the radio is approximately 1ms to calibrate an external 16MHz oscillator. We found that radios could either be self-clocked or externally clocked by the microcontroller. Either option requires a highly accurate oscillator on-board. We chose not to use an external high speed oscillator with the microcontroller and instead allow the microcontroller to sleep while the radio starts up, processes, or while sensors are acquiring data. For robustness, we chose to package a suite of sensors integrated onto the mote. Telos uses USB for power, programming, and communication with a host computer. Upon disconnection from USB, the mote runs on batteries that are isolated from the USB circuitry to prevent current leakage. An integrated antenna is robust against handling and environmental packaging; it is augmented with an external antenna connector for flexibility. Of importance to consumers, integrating a complete solution on to a single mote significantly reduces cost and eliminates the need for support hardware found in previous generations of motes (programming boards, interface boards, and sensor boards). All of the components in the system--the microcontroller, radio, flash, and sensors--operate down to 1.8V. By running at 1.8V for the life of a mote, it can double its lifetime by using half the power of running at 3.6V. With an LDO regulator and a small number of discrete components, the mote can utilize live almost twice as long on the exact same battery as previous mote generations. This collection of features has been integrated to create the highest data rate, lowest power mote to date. It lowers the barrier of entry to using wireless sensor networks for commercial applications by lowering cost, using standards such as USB and IEEE 802.15.4, and leveraging open source tools such as gcc and TinyOS [8]. References: [1] Chipcon AS. CC2420 2.4GHz IEEE 802.15.4 compliant RF Transceiver. November 2003. http://www.chipcon.com [2] Jason Hill, Robert Szewczyk, Alec Woo, Seth Hollar, David Culler, Kristofer Pister. System Architecture Directions for Networked Sensors. ASPLOS 2000: 93-104. [3] Jason Hill, David Culler. Mica: A Wireless Platform for Deeply Embedded Networks. IEEE Micro 22(6): 12-24 (2002). [4] Jason Hill. System Architecture for Wireless Sensor Networks. PhD Thesis, University of California, Berkeley, 2003. [5] IEEE Standard for Information Technology: 802.15.4: Wireless Medium Access Control and Physical Layer Specifications for Low-Rate Wireless Personal Area Networks. 2003. [6] Alan Mainwaring, Joseph Polastre, Robert Szewczyk, David Culler, John Anderson. Wireless sensor networks for habitat monitoring. WSNA 2002: 88-97. [7] Texas Instruments. MSP430 Microcontroller: F1611 User's Guide. 2004. http://www.ti.com. [8] TinyOS: An Operating System for the Wireless Sensor Regime. http://www.tinyos.net.