MATLAB/Simulink-Based Online CNN-LSTM Framework for EDFA Gain, OSNR, and ASE Noise Prediction with AI Decision Support
DOI:
https://doi.org/10.65405/edtrfz05Keywords:
EDFA, CNN-LSTM, MATLAB, Simulink, OSNR, ASE noise, optical communication, AI decision supportAbstract
This paper reports a MATLAB/Simulink simulation study for predicting Erbium-Doped Fiber Amplifier (EDFA) gain, optical signal-to-noise ratio (OSNR), and amplified spontaneous emission (ASE) noise by using a hybrid convolutional neural network and long short-term memory (CNN-LSTM) model. The work was implemented in MATLAB R2022b. A dynamic data set was generated in MATLAB from six operating variables: input optical power, pump power, wavelength, temperature, active channel count, and span length. The trained network was saved and deployed in Simulink using a Predict block. Two Simulink models were built: an offline validation model using stored test signals and an online simulation model using runtime source blocks. The word online is used only in the Simulink simulation sense; the model is not connected to physical EDFA hardware, optical spectrum analyzers, or pump-control equipment. Output-specific evaluation produced RMSE values of 1.4581 dB for gain, 2.7702 dB for OSNR, and 1.5373 dBm for ASE noise. A rule-based AI decision-support layer was also added to generate status, pump-action, and alarm indicators. The study demonstrates a complete software workflow that can be used as a preliminary platform before measured EDFA data or hardware-in-the-loop testing are introduced
Downloads
References
[1] C. R. Giles and E. Desurvire, “Modeling erbium-doped fiber amplifiers,” Journal of Lightwave Technology, vol. 9, no. 2, pp. 271–283, Feb. 1991.
[2] Y. Liu, X. Liu, L. Liu, Y. Zhang, M. Cai, L. Yu, and W. Hu, “Modeling EDFA gain: approaches and challenges,” Photonics, vol. 8, no. 10, Art. no. 417, 2021, doi: 10.3390/photonics8100417.
[3] Z. Wang, A. Yang, P. Guo, and P. He, “OSNR and nonlinear noise power estimation for optical fiber communication systems using LSTM based deep learning technique,” Optics Express, vol. 26, no. 16, pp. 21346–21357, 2018.
[4] F. Da Ros, U. C. de Moura, and M. P. Yankov, “Machine learning-based EDFA gain model generalizable to multiple physical devices,” in Proc. European Conference on Optical Communication (ECOC), 2020, doi: 10.1109/ECOC48923.2020.9333297.
[5] MathWorks, “Deep Learning Toolbox Documentation,” MathWorks, accessed Jun. 2026.
[6] MathWorks, “Predict responses using a trained deep learning network in Simulink,” MathWorks, accessed Jun. 2026.
[7] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997, doi: 10.1162/neco.1997.9.8.1735.
[8] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. Cambridge, MA, USA: MIT Press, 2016.
[9] S. Zhu, C. L. Gutterman, W. Mo, Y. Li, G. Zussman, and D. C. Kilper, “Machine learning based prediction of erbium-doped fiber WDM line amplifier gain spectra,” in Proc. European Conference on Optical Communication (ECOC), 2018.
[10] BenHusein, A. H., Shakrum, F. M., & Elgdiri, E. M. (2026). A Comprehensive Performance Evaluation of a Wi-Fi-Based Wireless Sensor Network Using Raspberry Pi and ESP8266 Under Multi-Rate Traffic Conditions. Al-Farooq Journal of Sciences, 2(3), 816-818.
