Exploiting Capture Effect to Provide Service Differentiation in Wireless LANs Alfandika Nyandoro ∗ and Mahbub Hassan ∗† ∗ School of Computer Science and Engineering University of New South Wales Kensington NSW 2052, Australia Email:{alfan, mahbub}@cse.unsw.edu.au † National ICT Australia Ltd Locked Bag 9031 Alexandria NSW 1435 Australia Abstract—We investigate the use of transmit power as a mechanism for service differentiation in contention based wireless Local Area Network(LANs). We use a dual transmit power scheme where the power levels are specially selected to guarantee capture whenever a collision occurs between a single high power frame and one or more low power frames. We employ the Distributed Coordination Function (DCF) of the IEEE’s 802.11 wireless LAN standard for our simulation experiments. The performance metrics considered are throughput and Medium Access Control (MAC) delay. We specifically investigate the level of differentiation achievable between high power and low power hosts given various ratios of high power to low power hosts. Our results show that differentiation is more distinct when the ratio of high power to low power hosts is small. It was also determined that in a cell with a fixed number of hosts capable of switching between the two power levels, there exists a threshold beyond which the MAC delay for the low power hosts actually decreases with increasing number of high power hosts. I. I NTRODUCTION In contention based access networks, such as ALOHA [1] and its various derivatives, the channel is shared by a number of hosts. Since channel access is contention based, if two or more hosts attempt to send frames in the same time slot, a collision occurs resulting in the loss of all affected packets. However, if the colliding frames are such that one of the colliding frames’ signal is sufficiently stronger than the aggregate strength of the rest, it is possible to correctly receive the former with the rest filtered out as noise. This is the so-called capture effect. Initial work on the capture effect (in the context of improving slotted ALOHA throughput) was first published by Metzner [2] more than two decades ago where it was observed that utilisation of ALOHA networks could be improved from 37% to 53% by simply dividing the transmitters into two groups, one group sending at high power and another sending at low power. Considerable work has been done on the capture effect since Metzner’s pioneering research. Most of the existing works investigate the use of capture to enhance overall system throughput and fairness e.g [3], [4]. Others investigate the fundamentals of various capture models, e.g. for indoor and outdoor environments [5], [6] or with respect to the modulation and coding techniques used by the transmitters, e.g. [7], [8]. In recent times, interest on the effect of capture in IEEE 802.11 [9] based networks has grown [10]–[13]. In [11], the fairness of a wireless LAN in the presence of the capture effect was investigated. A new capture model was also proposed since the investigation concluded that existing capture models did not agree with empirical data. In [13], the capture proba- bility in infrastructure wireless LANs was also investigated. In both [11] and [13], the contending hosts are of uniform trans- mit power; the difference in received signal strength (RSS) was due to the the near-far effect and multi-path propagation. In [12] the impact of heterogeneous transmit powers for hosts in an ad-hoc network was considered. The results indicated that the system suffered throughput degradation because high power hosts dominated the channel at the expense of low power hosts. In this paper, we explore the possibility of achieving ser- vice differentiation by deliberately creating and exploiting the capture effect in contention based wireless LANs. We specifically consider the Distribution Coordination Function (DCF) of the IEEE’s 802.11 standard. We note that although existing works acknowledge that capture results in unfairness on the channel, to our knowledge, none of the works have sought to leverage this unfairness in order to achieve service differentiation. Using transmit power for service differentiation appears attractive since no additional functionality is required at either the host or the access point. In addition, hosts are free to choose their transmit power, thus dovetailing well with the distributed nature of Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA), the protocol on which DCF is based. Our simulation study considers both throughput and MAC delay in a wireless LAN where the hosts can switch between two power levels. The rest of this paper is organised as follows. In Section II, the system model guiding our simulation study is presented, followed by a discussion of the results in Section III. We conclude the paper in Section IV with some notes on our ongoing efforts in this area. II. SYSTEM MODEL We simulate an ideal channel condition infrastructure wire- less LAN in which hosts exchange frames through the Access US Government work not protected by US copyright 1174 0-7803-8521-7/04/$20.00 © 2004 IEEE