In this study, the authors investigate average capacity of free space optics communication over Málaga atmospheric turbulence channel with pointing errors and path loss, for intensity modulated/direct detection (IM/DD) and heterodyne detection. Various algorithms which use adaptive transmission with both types of detection are considered, such as: optimal rate adaption (ORA), optimal power and rate adaption (OPRA), channel inversion with fixed rate (CIFR) and truncated channel inversion with fixed rate (TIFR). Analytical closed‐form expressions for channel capacities of ORA, OPRA and TIFR adaptive transmission are presented, and the authors prove that CIFR transmission is not feasible in the strict sense for the conditions considered. Obtained analytical results are numerically evaluated and graphically presented for different strengths of atmospheric turbulence (in weak, moderate and strong turbulence regime) for both types of detection (IM/DD and heterodyne), and for considered algorithms of adaptive transmission (ORA, OPRA and TIFR). The authors have developed expressions suitable for approximating high signal‐to‐noise ratio channel capacity, and they graphically present and compare the asymptotic approximations with the obtained analytical results for different strengths of turbulence for both types of detection. Also, obtained analytical results were confirmed by Monte‐Carlo simulations, and graphically compared for different strengths of turbulence regimes.
1 January 2018 in Facta universitatis - series: Electronics and Energetics
In this paper macrodiversity system consisting of two microdiversity SC (Selection Combiner) receivers and one macrodiversity SC receiver are analyzed. Independent κ-μ fading and correlated slow Gamma fading are present at the inputs to the microdiversity SC receivers. For this system model, analytical expression for the probability density of the signal at the output of the macrodiversity receiver SC, and the output capacity of the macrodiversity SC receiver are calculated. The obtained results are graphically presented to show the impact of Rician κ factor, the shading severity of the channel c, the number of clusters μ and correlation coefficient ρ on the probability density of the signal at the output of the macrodiversity system and channel capacity at the output of the macrodiversity system. Based on the obtained results it is possible to analyze the real behavior of the macrodiversity system in the presence of κ-μ fading.
1 November 2015 in Journal of Electrical Engineering
A system with macrodiversity selection combining (SC) receiver and for microdiversity equal gain combining (EGC) receivers is considered. Received signal is subjected, simultaneously to multipath fading and shadowing, resulting in signal envelope and signal power variation. Closed form expressions for moments of macrodiversity SC receiver output signal envelope are calculated. Numerical expressions are plotted to present the influences of Gamma shadowing severity and Nakagami-m severity on moments of proposed system output signal.
This paper considered the macrodiversity system with three micro diversity MRC receivers and one macrodiversity SC receiver. On the inputs present are fast k-μ fading and slow Gama fading which is correlated. For such macrodiversity system calculated are signal moments on its output. Achieved results are graphically presented. The influence of Rician factor, the depth of shading channel c and correlation coefficient ρ on signal moments was analysed.
In this paper, we investigate a wireless mobile communication system with a cooperative network operating over a mobile-to-mobile (M2M) fading channel in the presence of co-channel interference. Since there is no direct line of sight between the source (S) and the destination (D) due to various obstacles, the signal transmission takes place by using a relay (R). We modeled and represented the desired signal and co-channel interference by using the Nakagami-m fading distribution from the source (S) to the relay (R), as well as from the relay (R) to the destination (D). As the popularity of M2M communication has increased in 5G cellular networks, the performance analysis of the proposed system is of great importance. We derived closed-form expressions for the probability density function (PDF) and cumulative distribution function (CDF) of the signal-to-interference ratio (SIR) at the input of the destination station. Based on the cumulative distribution function (CDF), we also evaluated the outage probability (Pout) of the proposed communication system. In addition, we derived a closed-form approximate expression for the level crossing rate (LCR) by using the Laplace approximation formula for the three-fold integral. Validity of the derived theoretical results we approved by Monte Carlo simulations.
This paper introduces an adaptive A-semilogarithmic gradient quantization framework aimed at reducing memory overhead and computational complexity during the training of deep neural networks. The approach employs a semilogarithmic companding function parameterized by a dynamically adjusted scaling factor A, which evolves in response to the statistical properties of gradients throughout the training process. Two distinct quantization strategies are proposed and evaluated: The switching piecewise A-quantizer, which adaptively toggles between low-bit uniform and high-bit semilogarithmic quantization according to an exponentially weighted moving-average (EMA) estimate of gradient variance; and the hybrid A-quantizer, which statically partitions the gradient domain, applying uniform quantization in low-magnitude regions and semilogarithmic companding in high-magnitude regions. The proposed methods are empirically evaluated on both multilayer perceptron (MLP) and convolutional neural network (CNN) architectures using tabular and image-classification benchmarks, including DCCC, CIFAR-10, CIFAR-100, and ImageNet. Quantitative results demonstrate that both models achieve comparable classification accuracy to full-precision (FP32) baselines while significantly reducing gradient reconstruction error. Notably, the hybrid A-quantizer consistently yields better validation accuracy, reduced RMSE, and improved convergence behavior relative to its switching counterpart. These findings underscore the effectiveness of hybrid semilogarithmic quantization as a robust and efficient solution for training deep models in resource-constrained or bandwidth-limited environments, with strong potential for scalable deployment across diverse hardware platforms.