POWER MANAGEMENT STRATEGY

POWER MANAGEMENT STRATEGY

From the viewpoint of the load, all power management is based, in the final analysis, on three principles:

• Do not buy any unnecessary lights. That is to say, do not waste power on any unnecessary tasks. This is largely a function of the RF communication and sensor interface protocols to perform their tasks as efficiently as possible, by placing as little a load as possible on the hardware. An example of this is the trade-off between the energy costs of communication and computation. A typical wireless sensor network transmission and reception cycle may take 1 ms and require hardware that consumes 35 mW during that time, for a total energy consumption of 35 μJ — all to send perhaps 10 bytes of information, for an average energy consumption of 437.5 nJ/bit. (Due to fixed communication overhead, longer messages would be somewhat more efficient.) Computation on energy-efficient microcomputers requires on the order of 1–10 nJ/operation, however, so it is possible to do a significant amount of processing for the energy cost of one transmission. It is, therefore, energetically advantageous to perform a significant amount of raw data processing in the node itself, prior to transmission, if that processing either eliminates the need to transmit completely, or reduces the amount of data that is to be transmitted.

• Turn the lights off when you leave the room. In terms relevant to wireless sensor networks, do not waste power on any unused circuits. Turn off oscillators, pad drivers, in fact, all circuits when not in use. For CMOS logic circuits, this means to employ clock gating to stop the clocks to unused sections of logic; for low-voltage CMOS logic, where leakage current may be significant, it means employ multithreshold logic, so that the leakage may be reduced. For transceivers, this means to enable circuits in stages during warm-up, instead of all at once, and employ techniques to minimize warm-up time.

• Dim the lights when you are in the room. Use the minimum possible energy to complete the required task. For example, from Equation 8, the supply voltage and clock frequency of digital circuits may be dynamically varied over time to meet the application requirements — when the computational load is heavy, the supply voltage and clock frequency may be increased, then reduced when the computational load lightens. In RF transceivers, the transmit power may be reduced to the minimum necessary to complete the communication link; similarly, the RF amplifier in the receiver may be disabled and bypassed under strong signal conditions.

Viewing the system as a whole, however, the guiding principle is one of matching the (hopefully minimized) energy and power requirements of the load with the (hopefully maximized) energy and power available from the source. The better the source and load are matched, the less elaborate (and inefficient) the required power-conditioning circuits must be, and the more successful the overall design. One notes that the lower the power consumption requirements of the load, the more types of power sources, system designs, and applications will be compatible with it, and the greater its market flexibility. The load power consumption may then be used as a metric to estimate the economic viability of a given desing.