Large-Scale Synthetic Dataset for Deep Learning-Based TWDM-PON Parameter Optimization (1 Million Samples)

This dataset contains 1,000,000 synthetic samples representing the physical layer performance of a Time-and-Wavelength Division Multiplexed Passive Optical Network (TWDM-PON). The data was generated to support Deep Learning (DL) and Machine Learning (ML) research in optical network optimization, specifically for extending transmission reach and minimizing Bit Error Rate (BER).

Methodology: The dataset was created using a physics-based simulation environment that models the non-linear transfer functions of Single Mode Fiber (SMF). It incorporates both linear impairments (attenuation, chromatic dispersion) and non-linear impairments (Self-Phase Modulation/Kerr effect) based on standard optical link budget equations.

Data Structure: The dataset is provided as a CSV file (twdm_pon_synthetic_dataset.csv) with 16 columns:
- Input Features (11 Parameters):
- tx_power: Transmitter Power (8–15 dBm)
- distance: Link Distance (0–80 km)
- wavelength: Channel Wavelength (1534 nm / 1596 nm)
- fbg_reflectivity: Fiber Bragg Grating Reflectivity (0.85–0.99)
- laser_linewidth: Laser Linewidth (0.1–2.0 MHz)
- modulation_index: Modulation Index (0.5–0.95)
- fiber_dispersion: Chromatic Dispersion (15–18 ps/nm·km)
- amplifier_gain: Inline Amplifier Gain (0–6 dB)
- noise_figure: Amplifier Noise Figure (3–6 dB)
- extinction_ratio: Extinction Ratio (10–15 dB)
- apd_gain: Avalanche Photodiode Gain (5–15)

Target Metrics (5 Outputs):
- received_power: Optical Power at Receiver (dBm)
- q_factor: Signal Quality Factor (dB)
- ber: Bit Error Rate (calculated from Q-Factor)
- receiver_sensitivity: Required Sensitivity (dBm)
- power_budget: Total Link Power Budget (dB)

Potential Applications: This dataset is suitable for training regression models (DNN, CNN, Random Forest) to predict Quality of Transmission (QoT) or for reinforcement learning agents optimizing network configurations.