The most difficult constraints in the Smart Dust design are those regarding the minimum energy consumption necessary to drive the circuits and MEMS devices. The basic concepts involved in the structure of Smart Dust Mote are illustrated in figure-1. When fitting the entire mote within a 1mm cube volume, the energy density of the power supply is the primary issue. Current technology yields batteries with ~1J/mm􏰂 of energy and a high series resistance. Modern capacitors can achieve as much as ~10mJ/mm3 with a low series resistance. Series resistance affects the peak power that can be pulled from the source. In typical low power mixed-signal systems, most designers consider performance in terms of cycles, samples, or bits, maximising performance first and minimising power second. With the strict power constraints for Smart Dust, we have to consider performance in terms of Joules: given a cubic-millimeter battery, there is one Joule of energy to use. With the CCR, communication costs about 1nJ/bit, while sensing can be achieved at ~1nJ/sample. Modern processors, such as the StrongARM SA1100, can perform computations as low as ~1nJ/instruction. With these energy figures, one can make cost tradeoffs between the amount of computation, the amount of data transmitted and the sensor sampling frequency. However, by using a closer mapping of the application needs to the architecture and targeting ultra low energy from the start, we believe we can achieve orders of magnitude reduction in the energy cost per instruction.

In order to make realistic tradeoffs, a particular application scenario was chosen to present the design. We chose the case of military base monitoring wherein on the order of a thousand Smart Dust motes are deployed outside a base by a micro air vehicle to monitor vehicle movement. The motes can be used to determine when vehicles were moving, what type of vehicle it was, and possibly how fast it was traveling. The motes may contain sensors for vibration, sound, light, IR, temperature, and magnetization. CCRs will be used for transmission, so communication will only be between a base station and the motes, not between motes. A typical operation for this scenario would be to acquire data, store it for a day or two, and then upload the data after being interrogated with a laser.