The unprecedented cellular traffic growth caused by the ubiquitous mobile Internet access has motivated the search for solutions to cope with this explosive demand of connectivity. Mobile traffic offloading through Wi-Fi networks has emerged as an effective way of supplementing cellular services. This approach has been proven very effective to reduce the load of delay-tolerant traffic in cellular networks. Recent developments suggest that real-time traffic like VoIP could be also offloaded if Wi-Fi networks supported a reasonable number of concurrent calls. In this way, ubiquitous IEEE 802.11 networks operating in the 2.4 GHz ISM band could complement the cellular systems in indoor and rural environments. Wi-Fi networks are allocated 14 channels in this frequency band, thereof only 3 non-overlapped channels are allowed for Wi-Fi extended service sets in order to avoid interference. In this paper we investigate the use of assignment configurations with 4 partially-overlapped channels for VoIP transmissions, which may increase the network capacity in terms of concurrent users supported by Wi-Fi networks. Through computer simulations we demonstrate that 4-channel configurations are feasible at the expense of a tolerable degradation in quality of service for voice traffic with pedestrian mobility.

The capacity of a discrete noisy channel is achieved when the transmission source is properly matched to the channel, in the sense that the distribution of the transmitted bits maximizes the channel's mutual information. Traditional source matching assumes independency between the source and the channel, while the case of source dependent channel is usually ignored. In this article, the source dependant channel of Optical Code Division Multiple Access (OCDMA), dominated by Multiple Access Interference (MAI) is analyzed. It is shown that source dependency of MAI reverses the expected behavior of the OCDMA channel, and also that for a given value of channel condition, or mutual information, the number of concurrent users can be increased when the source is properly matched to the channel. A simple technique, which allows combining source matching with practical bit stream methods such as synchronization, error correction codes and the utilization of different detection methods, is also presented.

The fixed channel assignment problem in GSM networks has been commonly modelled as a constraint satisfaction problem with binary channel separation constraints (graph-colouring model). However, it has been observed that heuristics using direct estimations of carrier-to-interference ratio (CIR) have several advantages over those using binary constraints, being the economisation of spectrum the most important one. This is due to the fact that CIR constraints provide the possibility of taking more accurately into account different propagation phenomena that directly impact the channel assignment plans. In this paper, we present further improvement to an existing CIR-constraint heuristic algorithm for fixed channel assignment. We demonstrate that accounting for the change of the propagation coefficient at the break point in line-of-sight conditions (commonly seen in urban environments with rectangular grid street plans) might further economise spectrum. Results of practical examples are included

In this article we propose an algorithm for the optimal location of base stations in a mobile radio network. This task is a very complex combinatorial problem since many constraints and requirements must be satisfied. The proposed algorithm locates base stations in a traffic adapted manner, which means that base stations are concentrated inside traffic hotspots and are sparsely distributed in zones with lower traffic demand. It also assigns to all base stations practically the same traffic load, which enhances the efficiency of the network and eases the frequency assignment process.

A novel approach to power loading for orthogonal frequency-division multiplexing (OFDM) systems is proposed. The corresponding algorithm minimizes the aggregate bit-error rate, but appears to be rather complicated. Consequently, a much simpler, quasi-optimal loading algorithm is developed. The performance of various power loading strategies known in the literature are compared with the proposed quasi-optimal algorithm in the Rayleigh frequency-selective fading channel.

The correlated random process is represented as a solution of a stochastic differential first-order equation (SDE). The case of processes with a probability density function defined on a semi-infinite or finite range is considered. The limitation of the range of the simulated process requires modification of the specific structure of the SDE. The approach presented provides excellent results in modeling significant non-Gaussian processes with approximately an exponential correlation function. It is validated by direct numerical simulation of a uniformly distributed correlated process.