Abstract:
This dissertation demonstrates that it is possible to build a useful and complex sensor and communication platform in the cubic millimeter scale using optimizations at the circuit, computer architecture, component, and system-levels to contribute toward this goal.
The exponential size and power reductions in computation, communication, and sensing in recent years allows the integration of an autonomous wireless sensor node into volumes less than 16 mm^3. System architectures for several generations of such devices are presented.
One of the key components of wireless sensor nodes is the microcontroller that operates the system and provides data and power management. Three designs for ultra-lowenergy microcontrollers have been developed. The first provides a basis for ideal microcontroller calculations. The second features a hardware support for common sensor node functions to reduce the power consumption–with general instructions it consumes less than 14 pJ/instruction. The third design is a new reconfigurable datapath architecture with data-driven power cycling that has the potential to reduce energy consumption by orders of magnitude.
At a lower level, ultra-low energy circuit techniques are presented. Some unique items include a photovoltaic back-biased leakage reduction scheme, a 100 nW PTAT, and an 8 MHz, 60 nW integrated oscillator with a variable number of ring stages, wide current/frequency trimming, and nearly instantaneous power cycling.
Other system components are discussed also, including optical communication, microbatteries, and a micro radioisotope thermoelectric generator.
Publication date:
May 31, 2003
Publication type:
Ph.D. Dissertation
Citation:
Warneke, B. A. (2003). Ultra-low Energy Architectures and Circuits for Cubic Millimeter Distributed Wireless Sensor Networks. United States: University of California, Berkeley.