Carrier Recovery Techniques for Semiconductor Laser Frequency Noise for 28 Gbd DP-16QAM

Laser Rate Equation-Based Filtering for Carrier Recovery in Characterization and Communication

“…For coherent communication, the relevant laser parameters may be known in advance, but the corresponding state transition matrix may not yield a stable filter. As a general principle, decreasing K will both incur a larger phase noise penalty for a given Lorentzian linewidth [14] and increase the left hand side of (11). Thus the more detrimental the non-Lorentzian lineshape of the laser is to system performance, the more likely it is to be necessary to specify A in such a way that the system model is not completely consistent with the expected FM noise spectrum.…”

“…The Lorentzian linewidth of the simulated spectrum is larger (around 300 MHz) than the Lorentzian linewidth of the emulated spectrum (10 MHz). To emulate phase noise, we modulated an external cavity laser with a 40 kHz Lorentzian linewidth using a LiNbO 3 phase modulator and an arbitrary waveform generator at 50 GS/s [14]. Measurement noise was added using a standard (EDFA and attenuator) noise loading stage, and the signal was detected using a coherent receiver with a local oscillator with a 40 kHz Lorentzian linewidth and a digital sampling oscilloscope at 80 GS/s.…”

“…Because of this, there has been great progress in recent years toward compact integrated devices with narrow linewidths [33], [34]. However, the distinctive FM noise spectrum of an integrated SCL can cause standard carrier phase recovery algorithms to fail, even when the 3 dB linewidth of the laser is theoretically narrow enough to allow penalty-free transmission [14]. When the laser lineshape is far from Lorentzian, the EKF formulation discussed in Section II can yield a dramatically superior performance compared to CPR algorithms based on Lorentzian models.…”

“…We tested filter performance for a variety of damping frequencies experimentally and in simulation [13], [14]. Phase noise sequences were emulated experimentally using the technique described in Section IV-A and [14], [35].…”

“…We focus on two different Bayesian tracking algorithms: the particle filter (PF) and the extended Kalman filter (EKF). We then apply these filters in two different experiments where other carrier recovery methods were not successful: a measurement of the FM noise spectrum of a photonic crystal (PhC) cavity laser with extremely low output power [11], [12] and demodulation of a 28 GBd DP-16 QAM signal from a semiconductor laser-based transmitter [13], [14].…”