Free Space Optical (FSO) Communications. Towards the Speeds of Wireline Networks

Transcription

1 Free Space Optical (FSO) Communications Towards the Speeds of Wireline Networks

2 FSO Basic Principle Connects using narrow beams two optical wireless transceivers in line-of-sight. Light is transmitted from an optical source (laser or LED) trough the atmosphere and received by a lens. Provides full-duplex (bi-directional) capability. 3 optical windows : 850 nm, 1300 nm, & 1550 nm. WDM can be used => 10 Gb/s (4x2.5 Gb/s) over 1 Km & 1.28 Tb/s (32x40 Gb/s) over 210 m.

3 Why FSO? License-free Cost-effective Behind windows Fast turn-around time Suitable for brown-field Very high bandwidth (similar to fiber) Narrow beam-widths (point-to-point) - Energy efficient - Immune to interference - High level of security

4 FSO Applications Initially used for secure military as well as space applications Commercial use: Last mile solution, optical fiber back-up, high data rate temporary links, cellular communication backhaul, etc

5 FSO Challenges & Solutions Additive noise (photo-detector) and background radiation (direct, scattered, and reflected sun light) => sensitive detectors + filters + heterodyne detection Free space path loss => limited range Atmospheric losses (rain, snow, fog, aerosol gases, smoke, low cloud, sand storms, etc ) => power control + mesh architecture + hybrid RF/FSO Atmospheric turbulences => space diversity Buildings swaying, motion, and vibrations => tracking systems

6 Commercial Deployment Vendor Wavelength Data Rate Range 10 db/km) MIMO Hybrid RF/FSO Price Range (USD) fsona (Canada) 1550nm Full Duplex with 2.5 Gbps 1 km No Yes RF: 150 Mbps (60 70 GHz) 8-12K LightPointe (USA) 850nm 1550nm Full Duplex with 1.25 Gbps 1.6 kms Yes (2 X 2) (4 X 4) Yes RF: 250 Mbps ( GHz) 11-19K RedLine (South- Africa) 850nm Full Duplex with 1.25 Gbps 0.9 kms Yes (4 X 4) Yes RF: 250 Mbps ( GHz) 15-24K

7 Deployment Example: Lasers for High-Speed Traders (CNN)

8 Characterization of the Scintillations Frequency flat fading channel Channel coherence time: 10 μs and 100 ms Turbulence strength depends on Rytov variance/number (i.e. distance and index of refraction structure) Turbulence regimes: Rytov number << 1 => Weak turbulence regime Rytov number >> 1 => Strong turbulence regime Statistical models: Weak turbulence: Rice-Lognormal or Gamma-Gamma (Generalized K) Strong turbulence: Exponential or Gamma-Gamma (Generalized K) More generalized models: Double Gamma-Gamma or Malaga

9 Pointing Errors Definition: Thermal expansion, dynamic wind loads, and weak earthquakes result in the building sway phenomenon that causes vibration of the transmitter and the receiver known as pointing error. Effect on Communication (ξ): These pointing errors may lead to an additional performance degradation and are a serious issue in urban areas, where the FSO equipments are placed on high-rise buildings. Model: The pointing error model developed and parameterized by ξ which is the ratio between the equivalent beam radius and the pointing error jitter can be: - With Pointing Error: ξ is any number between 0 through 7 - Without Pointing Error: ξ

10 Generalized Pointing Errors Model The general model reduces to special cases as follows No misalignment Rayleigh Single sided Gaussian Hoyt Rician

11 Generalized Pointing Errors Model The fraction of collected power at the receiver can be approximated by [Farid and Harilovic, IEEE/OSA JLT, 2007] with r = r = x 2 +y 2 is random

12 On-Going Research Directions Unified performance analysis accounting for type of detection, weak/strong scintillations, and pointing errors. Computation of ergodic capacity over generalized FSO fading channels High SNR and low SNR bounds and approximations Bounds and exact results for the capacity of diversity systems Accurate approximations Average probability of error computations over generalized FSO fading channels Differentially coherent vs. coherent system performance Asymptotic results (coding and diversity gains)

13 On-Going Research Directions: Ergodic Capacity Computation High SNR and Low SNR Results over FSO channels. Bounds on the Capacity of Selection Diversity Systems Exact Capacity Results for MRC and EGC Diversity Systems Approximate results using PDF approximation

14 On-Going Research Directions: Asymptotic Analysis of Ergodic Capacity Heterodyne Detection Unified SNR Statistics IM/DD Unified with irradiance I = I a I p

15 On-Going Research Directions: Ergodic Capacity Calculations under the Impact of Pointing Errors Asymptotic Ergodic Capacity Recall that the irradiance I = I a I p and SNR g is proportional to I r The asymptotic ergodic capacity can be obtained as [Yilmaz and Alouini, SPAWC2012] We need to find the moments of I a and then, compute derivatives.

16 On-Going Research Directions: Asymptotic Analysis of Ergodic Capacity Exact Closed-Form Moments I= I a I p = I R I L I p where I R, I L, and I P are independent random processes Unified Rician Moments

17 On-Going Research Directions: Asymptotic Analysis of Ergodic Capacity High SNR Asymptotic Results Low SNR

18 On-Going Research Directions: Asymptotic Analysis of Ergodic Capacity Asymptotic Results Figure: Ergodic capacity results for IM/DD technique and varying k at high SNR regime for RLN turbulence

19 On-Going Research Directions: Ergodic Capacity Calculations under the Impact of Pointing Errors Generalized Pointing Errors Model The fraction of collected power at the receiver can be approximated by [Farid and Harilovic, IEEE/OSA JLT, 2007] Such that r = r = x 2 +y 2 is Beckmann distributed RV So

20 On-Going Research Directions: Ergodic Capacity Calculations under the Impact of Pointing Errors Generalized Pointing Errors Model The general model reduces to special cases as follows No misalignment Rayleigh Single sided Gaussian Hoyt Rician

21 On-Going Research Directions: Ergodic Capacity Calculations under the Impact of Pointing Errors Asymptotic Ergodic Capacity The asymptotic ergodic capacity can be obtained as The moments of I a are known for both lognormal (LN) and Gamma- Gamma (ΓΓ). Then, the asymptotic capacity can be written as,

22 On-Going Research Directions: Ergodic Capacity Calculations under the impact of pointing errors Asymptotic Ergodic Capacity Figure: The ergodic capacity for composite log-normal channel (LN). (a) ξ x = 6.7 and ξ y = 5.1 (b) ξ x = 6.7 and ξ y = 0.9 (c) ξ x = 0.8 and ξ y = 0.9 Reference: H. Al-Quwaiee, H.- C. Yang, and M. -S. Alouini, On the Asymptotic Ergodic Capacity of FSO Links with Generalized Pointing Error Model, Submitted to ICC 15.

23 On-Going Research Directions: Average Probability of Error Computations SER Performance of MPSK and MDPSK Symbol error rate performance of MPSK and MDPSK over AWGN are given by [Pawula, TCOM 1999] and with

24 On-Going Research Directions: Average Probability of Error Computations Asymptotic SER Performance Comparison of MPSK and MDPSK Well known that MDPSK performs 3 db worse than MPSK in the Rayleigh fading channels when the SNR is asymptotically large [Ekanayake- TCOM 1990] Asymptotic SER performance of MDPSK with respect to MPSK over a fading channel with diversity order t+1 with and, ηπ h t sin 2 θ t+1 dθ. 0 Asymptotic SER performance of MDPSK with respect to MPSK over lognormal fading channel

25 On-Going Research Directions: Average Probability of Error Computations Comparison of SER for MPSK and MDPSK in Lognormal Fading Figure: Average SER of FSO using MPSK and MDPSK over weak turbulence Lognormal fading channels. Reference: X. Song, F. Yang, J. Chengand M. -S. Alouini, Asymptotic SER performance comparison of MPSK and MDPSK in fading channels, IEEE Wireless Comm Letters, 2014.

26 Concluding Remarks Summary and Next Steps?

27 Conclusion and Current Work Spectrum scarcity is becoming a reality This scarcity can be relieved through: Cognitive radio networks Extreme bandwidth communication systems Analytical and fast simulation results can be used to perform initial system level trade-offs On-going deployment and testing the capabilities of FSO systems in hot & humid desert climate conditions.

28 Thank You Questions?