A combination of the coherent detection and digital signal processing (DSP) deployed in spectrally-efficient optical fiber communications is being applied to free space optical (FSO) communications. The DSP enables adaptive frequency offset compensation between the transmitter and receiver laser diodes (LDs), and also the adaptive equalization of the non-ideal frequency response of the optical and electrical devices, by continuously updating values for the phase rotation angles and the coefficients of equalization. Due to atmospheric turbulence, the SNR can suddenly be reduced, so that the adaptation will diverge from its optimum. Then, even if the SNR recovers, it will take much longer than usual for the adaptation to re-converge because it will be starting from a diverged value. In this paper, we propose to control the calculations for updating the adaptation with a state machine based on a SNR estimated from the extracted clock amplitude. The updated values are periodically written into FIFO registers when the SNR is higher than the receiver threshold, and when the SNR degrades, we fix the values using the output of the FIFO registers. This prevents divergence of the adaptation and enables reuse of the values before divergence, taking into account the fact that estimating the SNR takes a finite time. We designed the proposed DSPs, and confirmed that these designs can be implemented in field programmable gate arrays (FPGA). In an offline experiment we evaluated this model of the proposed DSP design using a 2.5 Gbaud quadrature phase shift keying (QPSK) signal. The experimental results showed that the proportion of error free time is increased from 91% to 98% by the proposed technique.
|