All-digital Online Timing Recovery with a Kalman Filter

Kurzfassung

Free-space optical communication is a promising technology that enables secure and high throughput for inter-satellite and direct-to-Earth communication links and helps achieve global connectivity. One of the challenges for free-space optical communication systems with signals passing through the atmosphere is synchronizing the sender and receiver sampling times under strongly and quickly varying channel conditions, for example, due to scintillation caused by atmospheric turbulence. During periods of low signal quality, the sampling time should not be lost and in case it is, it should be recovered as soon as possible to limit data loss. This thesis aims to provide a novel approach for all-digital timing recovery aimed at free-space optical communication systems by combining the Lee algorithm, a feed-forward timing error detector, with a Kalman filter followed by an interpolator. The approach combines information from the channel conditions under which the timing error measurement was made with the estimate based on a physical model and previous measurements. The Kalman filter is first derived from the general Kalman filter equations. This is followed by a Python implementation and verification of the filter against a simulated typical free-space optical channel resulting into distortions of the timing error detection. The simulations show improved performance of the timing error detection during fading and during low signal-to-noise ratio when using a Kalman filter. The simulations are followed by an real-time implementation of the Kalman filter on an field programmable gate array (FPGA) as part of an all-digital timing recovery chain based on the Lee timing error detection algorithm. The FPGA is connected to an experimental test setup sending a single- and receiving a dual-polarized optical quadrature phase-shift keying (QPSK) signal at 2 GBaud with an optical fading testbench to validate timing recovery performance during signal fading. The results show that the Kalman filter implementation does not constrain system speed, uses only 4 \% of the total used resources for timing recovery and improves the bit-error-rate during fades when compared to using the Lee algorithm without the Kalman filter. The thesis therefore recommends using a Kalman filter when using the Lee algorithm in context of free-space optical communication with links passing through the atmosphere.