Underwater Optical Communication Using Software ... - IEEE Xplore

15 downloads 172 Views 134KB Size Report
William C. Cox, Jim A. Simpson, and John F. Muth. Dept. of Electrical and Computer Engineering. North Carolina State University. Raleigh, NC 27695-7914.
The 2011 Military Communications Conference - Track 5 - Communications and Network Systems

Underwater Optical Communication Using Software Defined Radio Over LED and Laser Based Links William C. Cox, Jim A. Simpson, and John F. Muth Dept. of Electrical and Computer Engineering North Carolina State University Raleigh, NC 27695-7914 Email: {wccox,jasimpson,muth}@ncsu.edu Abstract—Underwater optical communication is an attractive means to achieve high datarate, low latency, and covert communication between underwater vehicles or sensor nodes. We demonstrate the viability of using a software defined radio system to communicate at Mbps rates using LEDs and lasers underwater and examine the performance of BPSK and GMSK simplex and duplex links.

I. I NTRODUCTION Underwater communication between mobile ocean systems is of great interest to the scientific and military communities. The limited propagation distance of RF frequencies [1] and low datarate of acoustic communication leave optical communication as a viable alternative for low-latency, high datarate communication in our oceans. By taking advantage of the “blue green” optical window in ocean water, underwater optical communication systems can utilize low cost optical sources, like diode lasers [2] or LEDs [3] and detectors like PIN photodiodes and PMTs. In this paper we show that software defined radio provides a convenient way to utilize a variety of modulation and demodulation schemes to fully implement an underwater optical communication system. Most optical communication systems working in situ utilize LEDs and baseband modulation, such as On-Off Keying (OOK) [4]–[8]. Laser-based links have also been demonstrated in laboratory environments using baseband modulation [9], [10]. While this may be appropriate in many applications, the ability to use high order modulation formats to send more bits per symbol is advantageous, but requires more complex passband modulation schemes [11], [12]. Modulating the intensity of the optical carrier using software defined radio allows one to leverage commercial, off-the-shelf, radio technologies to implement these more complex schemes for underwater communication. In this work we show the viability of a software defined radio link using underwater optical communication and passband modulation formats such as binary phase shift keying (BPSK) and Gaussian minimum shift keying (GMSK) for sending Ethernet network traffic. The digitizer used in this software radio configuration limited the link datarates to approximately 4 Mbps. For this paper, since the interest was examining the link over a variety of This work was supported by the Ofice of Naval Research via NRL grant N00173-07-1-G904 and STTR N00014-07-M-0308 and by the National Science Foundation under grant ECCS-0636603.

978-1-4673-0081-0/11/$26.00 ©2011 IEEE

simulated water conditions, the data presented was modulated at 1 Mbps to ensure a more robust link at higher attenuation lengths. II. T HEORY 1) Optical Attenuation: The two sources of loss in the underwater channel are absorption and scattering. Both are characterized by a differential loss per unit distance. Absorption is the destruction of the photon when encountering water molecules or particulate. Scattering is the redirection of the photon from its original path. Scattering in ocean water is highly peaked in the forward direction and can also have significant backscattering. The maximum loss over a given distance, which assumes no scattered photons are ever received, is expressed by Beer’s Law [13], I(λ : r) = I(λ : 0)e−cr

(1)

where r is the path over which the light travels, I(λ : x) is the intensity at a given wavelength and distance, and c is the attenuation coefficient. c is the sum of the absorption and scattering coefficients. The exponent, cr, is called the “attenuation length” and is unitless. The ratio between the scattering and total attenuation, ω, is the albedo, given by ω = b/c

(2)

where b is the scattering coefficient in units of m−1 and c is the attenuation coefficient given in m−1 . The albedo is nearly 1 for highly scattering environments and ranges from 0.5 to 0.9 for natural waters [14]. 2) Passband vs. Baseband Modulation: Most current underwater optical communication systems utilize baseband modulation [4], [5], [8], [15], where information is communicated using two levels of the received optical intensity, typically on and off. Modulation schemes such as OOK or PulsePosition Modulation (PPM) depend on the ability to determine the two levels of optical intensity for demodulation [16]. This unipolar signal requires knowledge of the DC level to establish a threshold, since a bipolar signal is not available as it is in RF communication. An alternate method is to use Subcarrier Intensity Modulation (SIM), which still employs direct detection (DD) of the signal, but the DC component of the signal is not used for conveying information [17]. For our

2057

%LDV7

5)$PS *185DGLR

6RIWZDUH

$PSOLILHG6L 3KRWRGLRGH

5[/HQV

+DUGZDUH

2SWLFDO %LDV7

5)$PS

/('/DVHU

:DWHU7DQN

7[/HQV

8653 $PSOLILHG6L 3KRWRGLRGH

'&%ORFN

Fig. 1. time.

7[/HQV

8653 '&%ORFN

*185DGLR

/('/DVHU

5[/HQV

Block diagram illustration of duplex communication link. The simplex links were conducted with the same setup, but only one link operating at a

experiments, SIM/DD was utilized to show its feasibility for optical communication underwater. The power requirements of SIM/DD are generally higher than baseband modulation methods, but the ease of detection and ability to easily transmit many bits per symbol with high order modulation schemes make it an attractive option. III. E XPERIMENTAL S ETUP Several experiments were performed to compare and contrast various underwater optical communication techniques. All tests were conducted in a 3.66 meter long by 1.22 meters wide and tall water tank. One end of the tank was equipped with a large window while the opposing end was equipped with a 20 cm view-port. Both windows are constructed from 0.64 cm thick polycarbonate. The tank was filled with municipal water and was filtered using a conventional pool pump with a diatomaceous earth filter. The tank holds approximately 3,800 liters of water. During experimentation a separate circulation pump was used to keep the water moving and any particulate in suspension. In order to simulate the scattering effects of ocean water, a commercial antacid, Maalox (a mixture of aluminum and magnesium hydroxide), was added to the water due to its similar scattering phase function to natural waters [18]. While the scattering function is similar, it should be noted that Maalox produces less forward scattering than natural waters and also exhibits a higher scattering vs. absorption ratio (albedo) than natural waters. For optical links operating at very high attenuation lengths, or links that capture a large amount of scattered photons, Maalox experiments will yield a higher number of captured photons than natural waters. To precisely control the particulate matter, a medical syringe pump was filled with Maalox and used to add the scattering agent to the water at a fixed rate over time while experimental data was captured. Although this 3.66 m tank is relatively short, previous pointto-point links that we have tested in both this tank and a 7 m tank at Naval Air Station Patuxent River give us confidence

that, with knowledge of the water parameters, results can be extrapolated to longer ranges. Receiver/transmitter pairs were placed on each end of the tank. A transmitter consisted of an optical source, either an LED or a diode laser, and the receiver was a Thorlabs PDA36A amplified Si photodiode detector with a 2 inch collection lens. The receiver output was passed through a DC-block and into an Ettus Research USRP signal acquisition board. This performed digitization/sampling, demodulation, and basic filtering before streaming the digital samples to the PC for software processing in GNU Radio. On the transmit side, an output waveform was generated from the USRP board, amplified with an RF amplifier (TB-409-84+) and biased with a bias-t (ZFBT-6GW-FT) before the signal was connected to the optical source. A. Software Setup The open source software defined radio project, GNU Radio, was used for the experiments to perform the various radio transmission tasks on the data. Sample code included in the project was modified and used to perform bit-error rate measurements along with testing a bidirectional TCP/UDP/IP link. For the bit-error rate measurements, a string of 1’s was encoded with a 7-bit linear feedback shift register (LFSR) scrambler. At the receiver the data stream was decoded using the same code and the bit density of 1’s to 0’s was measured to determine the error rate. This was done to reduce the complexity of the receiver and eliminate the need for bit synchronization between receiver and transmitter. It should be noted that a single bit error would produce a string of three 0’s in the received, decoded, data stream. The bit-error rate was adjusted for this, but at very low SNRs, where bit errors may be adjacent, the BER can be underestimated. The signalto-noise ratio (SNR) was estimated using the squared mean of the received data divided by the measured variance of the data. This estimator tends to overestimate the SNR towards lower SNR values (below 7 dB), and for our data, saturates at 4 dB. A more sophisticated SNR estimator could be used to reduce the scatter of the estimates, but this method was judged to be

2058

adequate for determining a trend. For the BER tests, BPSK modulation was used while GMSK modulation was used for the network tests. In the future, various other modulation schemes can be tested and compared, while the modulation schemes presented here were used for convenience. B. Comparing Attenuation Coefficients A WetLabs C-Star transmissometer was used to measure the attenuation coefficient of the water at a specific wavelength. This instrument was calibrated to measure the attenuation coefficient only at 532 nm. Instead of using multiple instruments at various wavelengths, to measure the attenuation at other operating wavelengths, an experiment was conducted to determine the attenuation at the minimum wavelength. The results are shown in Fig. 2 which shows experimental data collected at 410 nm. A photodetector was used to measure the power of a 410 nm laser propagating through the water as Maalox was added. The detector had a 2.7◦ field-of-view and a 1/2” aperture. The theoretical Beer’s Law line at 532 nm, from Eq. 1, is also plotted based on the measured attenuation coefficient. An exponential fit yielded an attenuation coefficient at 410 nm that is approximately 1.57c532nm . At 448 nm, 474 nm, and at 516 nm the attenuation coefficient is 1.39c532nm , 1.27c532nm , and 1.07c532nm respectively. This allows us to roughly scale the performance of the various wavelengths based on the measured attenuation coefficient values at one wavelength. 0HDVXUHGLQWHQVLW\ #QP YVFYDOXH #QP  

$UELWUDU\LQWHQVLW\



θdiverg = dspot /fl

(3)

where θdiverg is the divergence angle in radians, dspot is the diameter of the emitter, and fl is the focal length of the lens. It is clear that the divergence can be minimized by maximizing the focal length of the lens or by choosing an emitter with a small spot size. For high-power LEDs, where the emitter size is large, focal length must be maximized, but at the expense of captured light by the transmitter optics. The geometric loss over distance is expressed by 2 τgeo = Drx /(Rθdiverg )2

(4)

where τgeo represents the fractional received optical power, and ranges from 0 to 1, Drx is the diameter of the receiver aperture, and R is the distance from the transmitter to the receiver. This equation assumes that the spot size at the receiver is larger than the aperture and the transmitter is within the receiver’s FOV. This shows that the geometric loss is proportional to the square of the divergence angle. The amount of light captured by the transmitter lens is proportional to the angle subtended by the lens. This angle is given by

0HDVXUHGGDWD QP H[SRQHQWLDOILW %HHU¶V/DZ QP

φlens = tan−1 (r/fl )

(5)

where r is the lens radius and φlens is the half-angle intercept that the lens makes with the LED. Simulations and experiments showed that the divergence of the beam should be minimized rather than minimizing the focal length of the transmitter lens of the LED to achieve maximum power at the receiver. Ideally, the transmitter lens would be maximized, while assuring the beam spot size at the receiver is less than or equal to the receiver aperture.





D. Optical Powers



  

the receiver due to the extremely wide transmit field of view (FOV). A large diameter aspheric lens can be used to reduce the transmit field of view, but the short focal length of the lens, coupled with the large emission area of the LED, still causes a large divergence of the beam. An approximate value of this divergence is expressed by











FYDOXH #QP





Fig. 2. Attenuation vs. attenuation coefficient, measured at 532 nm, of 410 nm light. Data shows 410 nm light is attenuated approximately 1.57 times more than 532 nm light in the laboratory experiments.

C. Transmitter/Receiver Optics The choice of transmitter optics for the LED plays an important role in the system performance. For the experiments in Section IV a Cree XR7090 high-power LED was used. Its angular distribution at full-width half-max is 110 degrees and is roughly cos(θ) in shape. While this allows the system to potentially receive an optical signal over a wide range of angular offsets, it severely limits the received optical power at

Each of the chosen optical transmitters operated at different optical powers. The LEDs were operated at a fixed voltage/current and the optical power was measured at these operating conditions. The diode laser operating power was chosen to be near the optical powers of the LEDs. The LED’s peak operating wavelengths were 516 nm, 474 nm, and 448 nm respectively, and were measured with a USB4000 spectrometer. The powers are shown in Table I. IV. R ESULTS Several sets of experiments were conducted to determine performance of various optical links. First, data was captured comparing the bit-error rate of the link versus the water quality. This was measured for a one-way link at various wavelengths. Secondly, a bidirectional link was established

2059

TABLE I T RANSMITTER OPTICAL POWERS . λ (nm) 516 474 448 410 *

Type LED LED LED Laser

bit synchronization, all of which contribute to a decreased system performance, but are necessary for real-time links. It is expected that careful system tuning could reduce this offset. The default GNU Radio values were chosen for the digital clock and data recovery loops in order to minimize experimental variables.

Power (mW) 13.2 38.4 39 27/31*

Two different diode lasers were used.

%(5YVFíYDOXH





that utilized both receivers and transmitters to establish an IP link between two computers. Total throughput and latency were measured versus the water quality for this link. While these measures are highly dependent on network configuration and the type of network traffic, it can be a good metric for obtaining a general idea of link performance. It should also be noted that the network link did not employ any forward error correction, which would significantly improve the link quality and transmission distance.

í



í

%(5



í



í

A. Simplex Link Results - LED



%(5YV615







QP/(' QPODVHU QP/(' 















FYDOXH P







í



Fig. 4. Bit error rate vs. attenuation coefficient for 474 nm LED (38 mW), 516 nm LED (13.2 mW), and 410 nm diode laser (27 mW) transmission. The c-values are scaled to match the transmission wavelengths.

í

%(5



í



QP/(' %36.WKHRUHWLFDO QP/(' QPODVHU

í











  (VW615 G%





Fig. 3. BER vs SNR for 474 nm and 516 nm LED link, and the 410 nm diode laser link.

To test the one-way link performance, a link was established and the BER and received SNR was measured while the turbidity of the water was increased. The average optical power at the receiver was also periodically measured, and is discussed in Section IV-E. Since the photodetector was AC-coupled to the receiver hardware, the average power was measured with a multimeter connected to the photodetector and in parallel with the receiver hardware. SNR measurements were made on the digitized, downsampled, data and therefore only represented the passband portion of the signal. The SNR calculations exhibited a large amount of fluctuation, which can be seen in Fig. 3 as the spread in the experimental points. However, the experimental results generally follow the curve of the theoretical BPSK performance within 3 dB. The theoretical curve does not take into account factors such as quantization noise, clock offset, and

The absolute system performance can be determined by looking at the BER versus the water turbidity. The water quality, as measured by a WetLabs C-Star transmissometer, measures the attenuation coefficient of the water at a fixed wavelength (532 nm for our instrument). Fig. 4 show the BER vs. attenuation coefficient performance of the links at 516 nm and 474 nm. The c values are scaled to match the respective wavelengths, as per the discussion in Section III-B. The 474 nm and 448 nm LEDs transmitted at 38 mW and 39 mW respectively, while the 516 nm LED was much lower, at 13.2 mW. For the LEDs, this optical power was measured at the output of the transmit lens. All three LEDs were operated at the same bias voltage. The difference in optical powers is striking, with the shorter wavelength LEDs providing a far greater electrical efficiency. B. Simplex Link Results - Laser Fig. 3 and Fig. 4 also show the simplex link results for the system operating with a 410 nm diode laser. The laser optical power was approximately 27 mW. The 410 nm wavelength experiences 57% more attenuation than the 532 nm wavelength, as discussed in Section III-B. The beam was colimated and was a few milimeters in diameter, while the LED beam was diverging and much larger in diameter than the receiver aperture. The increase in performance between 516 nm and 474 nm is primarily due to the extra optical power, while the performance gain from 474 nm to 410 nm is due to the reduction in geometric loss for the laser link.

2060

C. Duplex Results - LED Duplex network communication was achieved using the system illustrated in Fig. 1. A network link between two computers was established using LEDs of similar wavelengths (448 nm and 484 nm) and a link using two different wavelengths (516 nm and 474 nm). For each link, network throughput and round-trip latency were measured versus the water quality. The network link did not employ any error correction, so the performance drops dramatically as soon as the SNR was low enough to produce errors in successive packets. Network performance was measured using the Netperf program, which utilizes a server/client architecture running on the end PCs. Throughput was measured for both the incoming and outgoing link using packets of approximately 1500 bytes, while latency was measured by the round-trip frequency of small (70 byte) ack/nack packets. 7KURXJKSXW 0ESV YVFYDOXH

7KURXJKSXW 0ESV

URXQGíWULSODWHQF\ PV

     



  FYDOXH P  DYHUDJH





attenuation coefficient axis scaled to match each wavelength, while the latency graph, since it is the round-trip latency, is the average of the two wavelengths used in the experiment.

 

D. Duplex Results - Laser

 







  FYDOXH P





7KURXJKSXW 0ESV

 



QPQPí([S QPQPí([S

Fig. 6. Round-trip latency of LED links. One link uses 516 nm/474 nm LEDs, while the other uses 448 nm/474 nm LEDs. c value shown is the average of the attenuation coefficient for the wavelengths of the LEDs.



2XWJRLQJ QP í([S ,QFRPLQJ QP í([S 2XWJRLQJ QP í([S ,QFRPLQJ QP í([S



URXQGíWULSODWHQF\YVFíYDOXH





Fig. 5. Incoming and outgoing throughput for LED links. One link uses 516 nm/474 nm LEDs, while the other uses 448 nm/474 nm LEDs. c value is scaled to match each wavelength.



    

URXQGíWULSODWHQF\ PV

Fig. 5 and Fig. 6 show the network performance of a link using two LEDs with 516 nm and 474 nm wavelengths and a link using 448 nm and 474 nm wavelengths. Due to the lower optical power of the 516 nm LED, the incoming link throughput drops before the outgoing throughput, which operates with the 474 nm LED. It is unclear to what extent the limited throughput of the incoming link affects the throughput of the outgoing link, however in most circumstances, the aggregate performance is what is critical. In order to compensate for the different wavelengths and optical powers of the 516 nm and 474 nm LEDs, a similar duplex link was established using a 474 nm and 448 nm LED. These wavelengths are close enough to have similar optical attenuation in the water, but would still be able to be wavelength filtered if necessary. For these test, the two links were spatially separated by 10 cm or more and no optical filters were employed. Fig. 5 and Fig. 6 also show the ethernet performance of this link. The throughput graphs have the

7KURXJKSXW 0ESV  /DWHQF\ PV YVFYDOXH



2XWJRLQJ QP ODVHU ,QFRPLQJ QP ODVHU 



  FYDOXH P





  FYDOXH P







 

/DWHQF\

   





Fig. 7. Incoming and outgoing throughput for 410 nm laser link along with round-trip latency. c value is scaled to match wavelength.

Fig. 7 shows the network results for the laser-based ethernet link. Two 410 nm lasers were used, with approximately 30 mW of optical power each. Alignment, power, and detector

2061

differences contribute to the 0.1 m−1 difference between the incoming and outgoing performance. Round-trip latency is plotted along with the system throughput. E. Received Power For each of the experiments, the received signal power was measured at the detector. This was plotted against the measured attenuation coefficient at 532 nm using the transmissometer. Several factors caused the actual signal power to deviate from the predicted Beer’s Law-type loss explained in Section II-1. These include the divergence of the beam, the beam diameter, the receiver FOV, the receiver aperture, and the transmission wavelength. For the received signal powers an exponential fit line was applied and a deviation factor, α was calculated. The fit took the form Prx = ae−cα3.66

(6)

where a is the initial power estimate, 3.66 is the transmission length in meters, c is the measured attenuation coefficient at 532 nm, and α is the system dependent deviation from the measured attenuation coefficient. In other words, the system factors will cause the light to attenuate at a rate different than the rate predicted by the measured attenuation coefficient. The α measurements for the various experiments are presented in Table II. The α value estimates from the linear fit described in Section III-B is also listed for each wavelength. It can be seen that the experimental results, due to system factors like the large receiver aperture, exhibit less loss than predicted. TABLE II FACTORS SHOWING SYSTEM DEVIATION FROM MEASURED ATTENUATION COEFFICIENT AT 532 NM , ALONG WITH THE ESTIMATED ATTENUATION COEFFICIENT FOR EACH WAVELENGTH . λ nm 516 474 448 410

Type LED LED LED Laser

αmeas. 0.85 0.95 1.10 1.48

αest. 1.07 1.27 1.39 1.57

V. C ONCLUSION Underwater optical communication provides the potential for underwater observation to be increased, by allowing underwater vehicles and sensor nodes to gather and transmit larger amounts of data than they are currently able using acoustic communication. The high data rates and low latency make it feasible to establish network connections between these systems for retrieving or downloading large datasets. While many engineering challenges remain, such as how to acquire a link in a timely manner and maintain pointing and tracking, we have demonstrated two types of links using LEDs or lasers that enable network connectivity using a minimum amount of off-the-shelf hardware. By using the GNU Radio software and low-cost RF hardware, several types of optical links were demonstrated. A series of LED links were demonstrated that can illustrate system performance where pointing and tracking may be

eased due to the large divergence of the beam. Additionally, a laser-based link was demonstrated to show the increased performance of such a system. By using a software defined radio system, future improvements can be easy made in software to test various modulation formats or digital filtering. R EFERENCES [1] X. Che, I. Wells, G. Dickers, P. Kear, and X. Gong, “Re-evaluation of RF electromagnetic communication in underwater sensor networks,” Communications Magazine, IEEE, vol. 48, no. 12, pp. 143–151, 2010. [2] W. C. Cox, “A 1 Mbps Underwater Communication System Using a 405 nm Laser Diode and Photomultiplier Tube,” M.S. thesis, North Carolina State University, Raleigh, NC, 2007. [3] J. Simpson, “A 1 Mbps Underwater Communications System using LEDs and Photodiodes with Signal Processing Capability,” Ph.D. dissertation, NC State University, 2008. [4] J. A. Simpson, W. C. Cox, J. R. Krier, B. Cochenour, B. L. Hughes, and J. F. Muth, “5 Mbps Optical Wireless Communication with Error Correction Coding for Underwater Sensor Nodes,” in Proc. OCEANS Conf. 2010, Seattle, WA, 2010, pp. 6–9. [5] N. Farr, J. Ware, C. Pontbriand, and T. Hammar, “Optical communication system expands CORK seafloor observatory’s bandwidth,” in Proc. OCEANS Conf. 2010, Seattle, WA, 2010. [6] C. Pontbriand, N. Farr, J. Ware, J. Preisig, and H. Popenoe, “Diffuse high-bandwidth optical communications,” in Proc. OCEANS Conf. 2008. Quebec, Canada: IEEE, 2008, pp. 1–4. [7] M. Doniec, I. Vasilescu, M. Chitre, C. Detweiler, M. Hoffmann-Kuhnt, and D. Rus, “AquaOptical: A lightweight device for high-rate longrange underwater point-to-point communication,” in OCEANS 2009, MTS/IEEE Biloxi-Marine Technology for Our Future: Global and Local Challenges, vol. 44, no. 4. Biloxi, MS: IEEE, 2009, pp. 1–6. [8] M. Doniec and D. Rus, “BiDirectional Optical Communication with AquaOptical II,” in Embedded Networked Sensor Systems, Zurich, Switzerland, 2010. [9] F. Hanson and S. Radic, “High bandwidth underwater optical communication.” Applied Optics, vol. 47, no. 2, pp. 277–83, Jan. 2008. [10] M. Chen, S. Zhou, and T. Li, “The implementation of PPM in underwater laser communication system,” in Communications, Circuits and Systems Proceedings, 2006 International Conference on, vol. 3. Guilin: IEEE, 2006, pp. 1901–1903. [11] W. C. Cox, K. F. Gray, J. A. Simpson, B. Cochenour, B. L. Hughes, and J. F. Muth, “A MEMS Blue / Green Retroreflecting Modulator for Underwater Optical Communications,” in Proc. OCEANS Conf. 2010, Seattle, WA, 2010. [12] L. Mullen, A. Laux, and B. Cochenour, “Propagation of modulated light in water: implications for imaging and communications systems.” Applied Optics, vol. 48, no. 14, pp. 2607–12, May 2009. [13] C. D. Mobley, Light and Water : radiative transfer in natural waters. San Diego: Academic Press, 1994. [14] S. Karp, R. M. Gagliardi, S. E. Moran, and L. B. Stotts, Optical channels: fibers, clouds, water, and the atmosphere. New York: Plenum Press, 1988, vol. 424. [15] D. Anguita and D. Brizzolara, “Optical wireless communication for underwater Wireless Sensor Networks: Hardware modules and circuits design and implementation,” in Proc. OCEANS Conf. 2010, Seattle, WA, 2010. [16] J. Li, J. Q. Liu, and D. P. Taylor, “Optical Communication Using Subcarrier PSK Intensity Modulation Through Atmospheric Turbulence Channels,” IEEE Transactions on Communications, vol. 55, no. 8, pp. 1598–1606, 2007. [17] W. O. Popoola and Z. Ghassemlooy, “BPSK Subcarrier Intensity Modulated Free-Space Optical Communications in Atmospheric Turbulence,” Journal of Lightwave Technology, vol. 27, no. 8, pp. 967–973, Apr. 2009. [18] A. Laux, R. Billmers, L. Mullen, B. Concannon, J. Davis, J. Prentice, and V. Contarino, “The a, b, cs of oceanographic lidar predictions: a significant step toward closing the loop between theory and experiment,” Journal of Modern Optics, vol. 49, no. 3, pp. 439–451, Mar. 2002.

2062