From fibers to satellites: lessons to learn and pitfalls to avoid when optical communications move to long distance free space

Abstract

The paper summarizes the recent investigation on feasibility of adapting state-of-the-art coherent fiber-optics (FO) systems for Free Space Optical (FSO) scenarios. This investigation is critically dependent on the intertwined aspects of architecture, as well as device and propagation impairments (including the channel) appearing in the aforementioned systems. Towards this, the work identified the key system differences between the two systems. Particularly, the FSO channel model was investigated, impact of atmospheric turbulence on FSO was discussed and a channel series was generated. Subsequently, relevant FO techniques including coherent detection, wavelength division multiplexing and Time-Frequency packing (TFP) were reviewed. Another departure from FSO works was the emphasis on coherent reception; receiver architectures and diversity schemes were first investigated. The former strived to make fair comparison amongst the receivers considering the diverse nature of perturbation added, while the latter indicated gain in performance through increase of diversity order (2-4 dB gain). An immediate conclusion is a suggestion on adaptation of wavelength diversity when coherent receivers . The investigation also evaluated the capacity and outage of fast and slow fading channels with parameters motivated by the channel modelling work. The shaping gain was evaluated and an LDPC code design example was provided for FSO downlinks. Finally, TFP enabled a remarkable performance gain when applied to coherent detection schemes, but only marginal with direct detection. The paper concludes by pointing to the next steps that build on this investigation and the need to corroborate with measurements.

An Information-Theoretic Comparison Between Coherent and IM/DD Transmissions for Free Space Optical Communications

“…which depends uniquely on σ 2 N and not on µ N . A lognormal distribution for the scintillation parameter H is considered accurate at low to moderate scintillation indices values, e.g., when S is in the order of 0.1, a value corresponding to a 10 • elevation for a satellite transmitting at 847 nm lightwave carrier, or to a 18 • elevation for a satellite transmitting in the 'third window' at 1550 nm [10]. When more severe atmospheric conditions imply larger values of S, e.g., in the order of 1.0, then other distributions like the gamma-gamma or the K-distributions are considered more accurate [9].…”

“…In (11), B is the channel bandwidth, measured in bits per second. Alternatively, one can use the spectral efficiency η = C/B [bits/s/Hz], which can account for time-frequency packing formats [10]. In (11), the SNR is evaluated for a given value of the scintillation, hence it corresponds to either (2) or ( 5) for the two system models considered here, while the effective received SNR is its average, e.g., for a coherent system…”

“…In (12), we average over the conditional mutual information for a given scintillation value H = h and find the input distribution P (X) which maximizes the average rate expression. As it is known, these rates are upper bounded by the actual capacity expression in (10). Resorting to a uniform distribution of the constellation points may yield a further penalty with respect to capacity.…”

“…The adoption of coherent transceivers for FSO communications opens up the field of investigation to similar analyses, performed on a different channel model, where a general complex optical field, i.e., belonging to C and with signed components, carries the information [10], enlarging the signal space by a factor of two. A further doubling of the signal space is provided by the possibility to exploit polarization division multiplexing, with coherent transceivers, thus yielding a factor of four for the overall enlargement of the signal space.…”