Classes of Nodes with Different Power Amplifiers and their Influence in Wireless Multi-hop Networks 1 Martin Kubisch, Holger Karl, Adam Wolisz Telecommunication Networks Group Technische Universit¨ at Berlin {kubisch|hkarl|awo}@ieee.org Abstract: Optimizing wireless network operations can be achieved by transmission power control of wireless nodes, which is mostly performed by regulating the final power amplifier. Depending on the desired transmission power level, cur- rently popular amplifiers work at different levels of effi- ciency: highest in case of maximum output power, lower in case of reduced output power. As the power amplifier is the major power sink in most common network interfaces, op- erating at high efficiency is desirable. Ideally, a node would operate using a power amplifier which is tuned to provide its highest efficiency at the node’s (single or typically used) transmission power level. Our idea is to use specialization of nodes to allow for ef- ficient short-range communication using (mostly) ”short- range” specialists – nodes where the power amplifier is op- timized for low output power values – and to use the cor- responding ”long-range specialists” for the long-distance communication in one network. However, the node deploy- ment strategy should stay the same and all nodes have to be able to provide all power levels, albeit at different costs. We discuss the implementation and performance char- acteristics of such heterogenous networks in this paper. The results show a potential reduction in energy consumption to less than 65 % compared to classical networks with uni- form amplifiers. We characterize the optimum ratio of non- heterogenous nodes for a given network density with differ- ent types of node distributions and show how missing the optimum ratio will impact the energy consumption of the network. Keywords: Multi-hop wireless networks, power con- trol, sensor network. 1. Introduction Controlling the transmission power of wireless de- vices has been shown to be beneficial from topological and capacity perspectives. Since the transmission power output also influences the power consumption of a node via the power amplifier, this is also attractive from an energy efficiency point of view. This prospect, however, has to be constrasted with the actual hardware proper- ties. Today’s popular power amplifiers used in wireless network cards are designed to have the highest power efficiency at the maximum output power. When the out- put power is reduced, the power efficiency of the am- plifier decreases, i.e., the power consumed by an ampli- fier does not reduce with the same ratio as the output power decreases. As an example: the RF2155 power amplifier [12], which is designed for applications in the 915 MHz ISM band, has four different output power lev- els to perform power control for wireless network cards. 1 This work is funded by the German Ministry of Education and Research (BMBF) under the project AVM The power efficiency of these power levels range from 54 % efficiency for the highest level of output power to 1 % efficiency for the lowest level of output power. Power amplifiers are usually chosen according to the maximum range one wants to overcome. Making use of power control, assuming a desired packet error rate (PER) as well as other parameters of the receiver char- acteristic, leads to a requirement on the needed trans- mission power necessary to overcome the maximal dis- tance. In realistic setups, however, the distances between communicating nodes are variable. Thus, the usage of the highest output power is rarely necessary, even when direct communication between the sender and receiver is possible. Further options for using lower transmis- sion power appear when (as in IEEE 802.11) the cod- ing/modulation might be dynamically adjusted (slower data rates require lower transmission power to achieve the same PER). The desired and beneficial reduction of the output power does, unfortunately enough, not lead to a pro- portional reduction of power used to drive the amplifier, as the amplifier is moving into a less efficient operation range. But, in fact, there is no physical rule that man- dates that power amplifiers have the highest efficiency at the highest output power. A practical example for a dif- ferent amplifier design is given by CRIPPS [15], which presents amplifiers which are developed with efficiency enhancement techniques in mind. The presented “Do- herty” amplifier is an example of an adaptation of power efficiency. It has the highest power efficiency at a power level 6 dB less than the maximum. Obviously, using such amplifiers with shifted power efficiency would pay off if — mostly — lower output power is applied. This can be the case even when the costs for using the higher power level is un-proportionally high, but the higher power is rarely used. In an idealized situation, for each node in the network its desired transmission power would be determined and then a transceiver optimized for this transmission power would be chosen. Clearly, this is not possible, both from a deployment and a hardware point of view. Dynami- cally reconfigurable hardware might allow such an ap- proach, but such technology is not available. Therefore, only a discrete set of different transceivers can be assumed, and no control over their placement shall be assumed. We constrain ourselves to a setup where nodes can have one of two different types of am- plifiers. All nodes can provide the same two power lev- els (low and high), but operate with high efficiency only at one level. Consequently, nodes are “short range” or “long range specialists.” In Proc. of European Wireless Conference (EW), Volume 2, pp. 749 -- 755, Nicosia, Cyprus, April 2005 VDE/ITG.