Modeling and Performance Analysis of MDM−WDM FSO Link Using DP-QPSK Modulation Under Real Weather Conditions

Kaur, Tanmeet, Kaur, Sanmukh ORCID: https://orcid.org/0000-0002-1750-5684 and Ijaz, Muhammad ORCID: https://orcid.org/0000-0002-0050-9435 (2025) Modeling and Performance Analysis of MDM−WDM FSO Link Using DP-QPSK Modulation Under Real Weather Conditions. Telecom, 6 (2). 29. ISSN 2673-4001

Preview

Available under License Creative Commons Attribution. Download (6MB) | Preview

Abstract

Free space optics (FSOs) is an emerging technology offering solutions for secure and high data rate transmission in dense urban areas, back haul link in telecommunication networks, and last mile access applications. It is important to investigate the performance of the FSO link as a result of aggregate attenuation caused by different weather conditions in a region. In the present work, empirical models have been derived in terms of visibility, considering fog, haze, and cloud conditions of diverse geographical regions of Delhi, Washington, London, and Cape Town. Mean square error (MSE) and goodness of fit (R squared) have been employed as measures for estimating model performance. The dual polarization-quadrature phase shift keying (DP-QPSK) modulation technique has been employed with hybrid mode and the wave division multiplexing (MDM-WDM) scheme for analyzing the performance of the FSO link with two Laguerre Gaussian modes (LG00 and LG 01) at 5 different wavelengths from 1550 nm to 1554 nm. The performance of the system has been analyzed in terms of received power and signal to noise ratio with respect to the transmission range of the link. Minimum received power and SNR values of −52 dBm and −33 dB have been obtained over the observed transmission range as a result of multiple impairments. Random forest (RF), k-nearest neighbors (KNN), multi-layer perceptron (MLP), gradient boosting (GB), and machine learning (ML) techniques have also been employed for estimating the SNR of the received signal. The maximum R squared (0.99) and minimum MSE (0.11), MAE (0.25), and RMSE (0.33) values have been reported in the case of the GB model, compared to other ML techniques, resulting in the best fit model.