QAM-DMT of Hybridly Integrated EAM-DFBLD Chip-on-Board With Adaptive Machine Learning Algorithm for 149.6-Gbit/s BtB and 138-Gbit/s 10-km-SMF Transmission

By using the generalized frequency division multiplexing (GFDM) and the machine learning algorithm for self-feedback amplitude/phase pre-distortion, the electro-absorption modulator integrated distributed-feedback semiconductor laser diode (EAM-DFBLD) is encoded with improved transmission quality and capacity. Both back-to-back (BtB) and 10-km single-mode fiber (SMF) transmissions are preliminarily demonstrated for inter-data-center applications. Optimizing the EAM-DFBLD transmitter at 1306.8 nm with a bias current of 100 mA and a bias voltage of −1.5 V can provide a 34.5-GHz bandwidth for 16-QAM discrete multitone (DMT) and GFDM formats. The parametric comparison reveals that the 16-QAM GFDM with
K
= 16 can improve its performance from BtB to 10-km SMF transmission with decoding BER slightly increasing from 1.44 × 10
⁻³
to 2.82 × 10
⁻³
. The EAM-DFBLD carried GFDM data also provides a better power penalty by -0.25 dB. This enables higher raw data rates of 149.6 Gbit/s under BtB case and 138 Gbit/s in 10-km SMF transmission link below the forward error correction (FEC) criterion. The GFDM format carried by the EAM-DFBLD transmitter thus becomes the more suitable candidate than the DMT to implement ultra-high-speed inter-data-center and data access link applications.