Voice Call Capacity Over Wireless Mesh Networks

The goal of this thesis is to understand the voice call carrying capacity of an IEEE 802.11b/e based ad hoc network. We begin with the modelling of conversational speech and define a six state semi-Markov voice model based on ITU-T P59 recommendation. We perform a theoretical analysis of the voice model and compare it with results obtained via simulations. Using a Java based IEEE 802.11 medium access layer simulator, we determine the upper-bound for the number of voice calls carried by an ad hoc network. We use a linear topology with the ideal carrier sensing range and evaluate the number of calls carried using packet loss and packet delay as metrics. We observe that, for one, two, three and four hop, 5.5 Mbps IEEE 802.11 wireless links have an upper-bound of eight, six, five, and three voice calls respectively. We then consider a carrier sensing range and a path loss model and compare them with the ideal case. We observe, after considering a carrier sensing range with path loss model, there is a reduction in the number of calls carried by the linear networks. One, two, three and four hop 5.5 Mbps IEEE 802.11 wireless links support eight, five, four, and two voice calls respectively, when a carrier sensing range and a path loss model is considered. We also find that by adopting packet dropping policies at the nodes, we improve the call carrying capacity and quality of service on the network. In our simulations of a two hop network in path loss conditions, we find that, by adopting a time delay based packet dropping policy at the nodes, the number of calls supported simultaneously increased from five to six. In a four hop linear network we find that by total packet loss is reduced by 20%, adopting a random packet dropping policy and by 50% adopting a time delay based packet dropping policy. Although there is no change in number of calls supported, load on the network is reduced