PIERS ONLINE, VOL. 5, NO. 1, 2009 65

Experimental Study of Atmospheric Turbulence Effects onRoFSO Communication Systems

W. Ni1, Y. Miyamoto1, K. Wakamori2, K. Kazaura2,M. Matsumoto2, T. Higashino3, K. Tsukamoto3, and S. Komaki3

1Hamamatsu Photonics K. K., 5000 Hirakuchi, Hamakita-kuHamamatsu-shi, 434-8601, Japan

2GITS/GITI, Waseda University, 1011 Okuboyama, NishitomidaHonjo-shi, 367-0035, Japan

3Division of EEIE, Graduate School of Engineering, Osaka University2-1 Yamada-oka, Suita-shi, 565-0871, Japan

Abstract— Many existing and emerging broadband wireless services benefit from the Radioon Fiber (RoF) technology offering high speed and large bandwidth over long distance at a lowattenuation. Although RoF approach works well in many cases, in fiber not feasible area, free-space optics (FSO) communication techniques have received renewed interest as alternative forRadio-Frequency (RF) signal transmission-concept referred to as Radio on FSO (RoFSO)-becausethey can offer similar capacity while being cost effective and secure access without extensiveinstallation of copper or fiber infrastructure. We are currently in the development stage ofthe advanced RoFSO system. On the other hand, atmospheric turbulence manifested as beamwander, intensity fluctuation and beam spreading has significant influence on the performanceof conventional FSO as well as RoFSO systems. In order to measure, characterize and quantifythe influence of atmospheric turbulence in our system deployment environment, we described inthis paper a setup Radio Frequency-FSO (RF-FSO) link to test the basic characteristics of RFsignal transmission under various atmospheric and weather conditions e.g., clear weather, clouds,rain, fog and snow. We have attempted to quantify the strength of atmospheric turbulenceby calculating the refractive-index structure parameter C2

n from scintillation index indicatingintensity fluctuation caused by atmospheric turbulence. Correlation analysis on carrier-to-noiseratio (CNR) and C2

n was also made and presented. In addition, a comparative measurement tocharacterize intensity scintillation and Angle-of-Arrival (AoA) variance has also been presented inthis paper. The experimentally derived data and results will be available for antenna design, linkmargin prediction and evaluation of performance criterion of future RoFSO systems in differentdeployment environments.

1. INTRODUCTION

In the future ubiquitous society, consumers will be provided with reliable and cost effective com-munication systems that can support anytime, anywhere, any media they want. As a result,wideband radio links will become more prevalent in communication systems. To satisfy this in-creasing demand, the high capacity optical networks should be integrated with the flexibility ofradio networks. The aforementioned wideband demands can be specifically met with fiber basedwireless access schemes such as RoF [1] by which high-speed multimedia can be easily supported.In areas where optical fiber or other broadband networking technologies deployment are not feasi-ble, FSO communication links can be utilized as an alternative technology because they can offercapacity similar to optical fiber. RoFSO [2] link can be used to reliably transport multiple RFsignals comprising of various kinds of communication and broadcasting signals.

We are developing an advanced dense wavelength division multiplexing (DWDM) RoFSO linkwhereby the basic antenna design concept will be similar to the full-optical FSO antenna [3]. Full-optical FSO links [4] in Figure 1(b) utilize seamless connection of free-space beam to single-modefiber (SMF) therefore eliminating the necessity of converting the transmitted signal from optical-to-electrical (O/E) or vice versa in conventional FSO as in. Figure 1(a). The use of 1550 nm asthe transmission wavelength for FSO systems is attractive for DWDM RoFSO system as it willbe compatible with long-haul fiber optic technologies such as wavelength division multiplexing anderbium doped fiber amplifiers [5, 6]. Atmospheric turbulence, i.e., refraction-index fluctuations infree space, will destroy light beam spatial coherence, cause phase front distortion, alter and redirectthe flow of energy in the beam, and finally lead irradiance and AoA fluctuations in the plane of

PIERS ONLINE, VOL. 5, NO. 1, 2009 66

the receiver aperture. This can be characterized statistically by the variance of optical intensityas depicted in Figure 1(c). Moreover, in full-optical FSO links, the free-space transmitted opticalbeam will be received and coupled directly onto the SMF interface with core diameter of only10µm, that is more difficult to ensure the stability of the RoFSO communication [7].

Electrical

signal

Optical fiber Optical fiber

(a)

Time

Time

O/E E/O Optical

source

PD Electrical

signal

Optical amplifier Control mechanism

(b)

Optical fiber Optical fiber

(c)

Figure 1: (a) Conventional FSO link; (b) full-optical FSO link; (c) intensity variance attributed to atmo-spheric turbulence.

2. MEASUREMENT AND ANALYSIS OF ATMOSPHERIC TURBULENCE

2.1. Measurement SetupIn order to study atmospheric turbulence impact on the quality of the envisaged RoFSO, RF-FSO antennas (commercial available Canobeam DT-170 antennas) have been used to set up anexperimental link over 1km distance between two campuses of Waseda University in Tokyo. Asdepicted in Figure 2, an RF signal generator (Agilent E4438C) is used to generate a W-CDMAsignal which is propagated through the atmosphere using the RF-FSO antenna installed on therooftop of Bldg. 14 at the Nishi-waseda campus. On the opposite side, the corresponding RF-FSO installed on the rooftop of Bldg. 55 at the Okubo campus along with a signal analyzer isused to measure the received optical power and calculate the carrier-to-noise ratio (CNR) in 1minute interval. An optical FSO antenna has been set to measure optical attenuation and intensityfluctuation simultaneously. And a weather meter was used to record temperature, rain rate, andother weather condition data. Signals from both of them are sampled and processed by the monitorpersonal-computer (PC).

Sig. Analyzer

Sig. Generator

Optical FSO antenna

Atmospheric

turbulence

1km

RF-FSO antenna

RF-FSO antenna Weather meter

Figure 2: Schematic of RF-FSO test link.

2.2. The Refractive-index Structure Parameter C2n

The refractive-index structure parameter C2n (m−2/3) is the parameter most commonly used to

describe the strength of atmospheric turbulence. According to Andrews et al. [8], C2n can be ap-

proximately calculated as following:

C2n ≈

σ21

1.23k7/6L11/6(1)

PIERS ONLINE, VOL. 5, NO. 1, 2009 67

where k = 2π/λ is wavenumber, L is link distance and σ21 is called the Rytov variance [8] which

can be obtained by

σ21 ≈ σ2

I =< I2 > − < I >2

< I >2(2)

where the quantity I denotes the measured irradiance of the optical wave and the angle brackets<> denote an ensemble average or, equivalently, a long-time average. In the RF-FSO test link,optical power received by the optical FSO antenna was sampled as in Figure 3(a). It can be seenthat intensity denoted by the monitor voltage V caused by atmospheric turbulence fluctuates morestrongly on clear day than rainy day.

6:00 12:00 18:00 24:0080

90

100

110

120

130

140

Car

rier

-to-

nois

e ra

tio

(dB

c)

Local time

minimum CNR

average CNR

0:000 500 1000 1500 2000 2500 30000.0

0.5

1.0

1.5

2.0

2.5

3.0

Mon

itor

outp

ut (

V)

Time (msec)

heavy rainy day

clear day

(a) (b)

Figure 3: (a) Measured intensity fluctuations; (b) minimum and average CNR (August 6, 2008).

Calculated C2n values nearly show the same diurnal cycle on all clear days due to similar con-

ditions, cloudless sky when daily temperatures have also been relatively higher as shown in Fig-ure 4(a). Around noon, maximum C2

n value of about 2 · 10−13 m-2/3 can be found as the refraction-index fluctuations are caused almost exclusively by temperature variations. When the air tem-perature is close to ground temperature, for example at sunrise and during the night, C2

n valuesare lower and even reach the minimum (about 3·10−15 m-2/3). For the month of July in 2008, itrained frequently even though the temperature was still higher. As presented in Figure 4(b), thecalculated C2

n values were almost the same during a whole day without any characteristic diurnalpattern identified The phenomenon was the similar in other three seasons: on rainy days, C2

n valuesare lower than the ones on clear days in the same season. The main reason for that is on rainy day,temperature stratification of the atmosphere was stable. Figure 4(c) gives the C2

n values on cleardays in December 2007 and July 2008. The typical diurnal cycle of C2

n was both found except thatthe C2

n values were generally higher in summer than those in winter. That is because solar radiationis usually weaker in winter season, thus the temperature is relatively lower as well as temperaturestratification. The phenomenon can be more clearly seen on rainy days in Figure 4(d). Therefore,when we make consideration for effect of atmospheric turbulence on the future RoFSO commu-nication link, we should pay more attention in summer season with clear weather condition, e.g.,system link margin should be designed to be able to tolerate the strongest strength of atmosphericturbulence at that time with C2

n value of 2·10−13 m−2/3 obtained from our experimental setup inthe same deployment environment.

2.3. Correlation Analysis on CNR

One-day result of the recorded CNR values from the RF-FSO test setup has been displayed inabove Figure 3(b). Both minimum and average CNR were generally above the desired minimumthreshold value of 110 dBc for our present W-CDMA signal transmission. The maximum varianceof the minimum CNR was found at noon period when the strength of atmospheric turbulence wasalso larger as stated in Section 2.2. We have made correlation analysis on C2

n values and CNRdeviation which is defined as the difference between average and minimum CNR at the same timepoint. On clear day in summer season, CNR deviations increased with the increasing C2

n values asshown in Figure 5(a). On rainy day, CNR deviations caused by the atmospheric turbulence werethought to be mostly not larger than 6 dB as seen in Figure 5(b) while other larger deviations maybecaused by heavy rain drops. This phenomenon can also be found in other three seasons. So, from

PIERS ONLINE, VOL. 5, NO. 1, 2009 68

3:00 6:00 9:00 12:00 15:00 18:00 21:00 24:001E-15

1E-14

1E-13

1E-12

Clear day in Aug 2008

Ref

ract

ive

ind

ex s

truct

ure

Cn

2 (

m-2

/3)

Local time0:00

(a)

3:00 6:00 9:00 12:00 15:00 18:00 21:00 24:001E-15

1E-14

1E-13

1E-12

Clear day in Dec 2007

Clear day in Jul 2008

Ref

ract

ive

ind

ex s

tru

ctu

re C

n

2(m

-2/3)

Local time

0:00

(b)

(c) (d)

3:00 6:00 9:00 12:00 15:00 18:00 21:00 24:001E-15

1E-14

1E-13

1E-12

Rainy day in Jul 2008

Ref

ract

ive

ind

ex s

tru

ctu

re C

n

2(m

-2/3)

Local time

0:00

3:00 6:00 9:00 12:00 15:00 18:00 21:00 24:001E-15

1E-14

1E-13

1E-12

Rainy day in Dec 2007

Rainy day in Jun 2008

Ref

ract

ive

ind

ex s

tru

ctu

re C

n

2(m

-2/3)

Local time

0:00

Figure 4: Calculation results of C2n values on: (a) clear day in summer; (b) rainy day in summer; (c) clear

day in winter and summer; (d) rainy day in winter and summer.

the relationship between CNR deviation and C2n values at corresponding time point, we can get

an estimation of redundancy for CNR deviation to mitigate the effect of atmospheric turbulence,e.g., threshold of CNR deviation should be not less than 10 dB in the setup RF-FSO link systemto withstand strongest atmospheric turbulence.

0 2 4 6 8 10 12 14 161E-15

1E-14

1E-13

1E-12

Clear day in Aug 2008

Fitted curve

Ref

ract

ive

ind

ex s

tru

ctu

re C

n

2(m

-2/3)

CNR deviation (dB)

0 2 4 6 8 10 12 14 161E-15

1E-14

1E-13

1E-12

Rainy day in Jul 2008

Ref

ract

ive

ind

ex s

tru

ctu

re C

n

2(m

-2/3)

CNR deviation (dB)

(a)

(b)

Figure 5: C2n values vs. CNR deviation: (a) on clear day; (b) on rainy day in summer of 2008.

2.4. AoA FluctuationsWe have also set up a FSO link, depicted in Figure 6, for studying turbulence-induced phasefluctuations on the propagating optical wave. The link distance is 300 m between two buildings inHamamatsu Photonics K. K. On the receiver side, optical beam propagating through atmosphere

PIERS ONLINE, VOL. 5, NO. 1, 2009 69

will be focused onto a position sensitive detector (PSD) inside one FSO antenna with aperturediameter of 0.04m. A/D conversion of signals from the two receiving antennas will be madeby Peripheral Interface Controller (PIC). The interested data including received optical power,scintillation index, average arrival angles and variance of AoA were sampled at 10 kHz and recodedin 15 seconds interval in PC, as shown in Figure 6. The AoA variance is a result of phase changeand can relate to C2

n according to [8, 9]:

< β2 >= 2.91C2nLD

−1/3R (3)

where L is link distance, here is 300 m and D is the receiver aperture of 0.04 m. The C2n values

can be calculated from the received intensity data of the FSO antenna with PD by applying (1)and (2). Comparison was made between the simulated AoA variance from (3) and experimentallyrecorded one from PSD. On clear day, AoA variances in both X and Y directions agreed well withthe calculated ones. And all of those variances fluctuated with the C2

n values. On rainy day asin Figure 7(b), the experimental AoA variances also agreed with the calculated ones and changedwith C2

n except for some large deviation points caused by large rain drops. And AoA variancechanges more severely with clear sunny weather than cloudy or rainy ones as shown in Figure 7.The largest deviation happens mostly at noon time on clear day. Therefore, as expected, AoAvariance fluctuates well with the strength of atmospheric turbulence. Moreover, it is helpful thatwe can make a better estimation of AoA variance as long as we know the strength of atmosphericturbulence, i.e., C2

n values in the deployment environment in preparation for RoFSO system design,e.g., setting suitable tracking characteristics.

FSO antenna with PSD

FSO antenna

300m

FSO antenna with PD

A/D

A/D

Figure 6: Angle of arrival measurement setup.

(a) (b)

3:00 6:00 9:00 12:00 15:00 18:00 21:00 24:001E-16

1E-14

1E-12

1E-10

1E-8

1E-6

X-direction AoA variance

Y-direction AoA variance

simulated AoA variance

Ang

le o

f ar

riv

al (

AoA

) v

aria

nce

Local time

Cn

2(m

-2/3)

0:00 3:00 6:00 9:00 12:00 15:00 18:00 21:00 24:001E-16

1E-14

1E-12

1E-10

1E-8

1E-6

X-direction AoA variance

Y-direction AoA variance

simulated AoA variance

An

gle

of

arri

val

(A

oA

) v

aria

nce

Local time

Cn

2(m

-2/3)

0:00

Figure 7: Simulated and experimental AoA variances: (a) on clear day (Mar 18th, 2008); (b) on rainy day(Mar 19th, 2008).

3. CONCLUSIONS

We have presented some preliminary experimental results with more emphasis on the effect ofatmospheric turbulence from an existing RF-FSO test bed which is significant in the design and

PIERS ONLINE, VOL. 5, NO. 1, 2009 70

evaluation of the advanced RoFSO system. The results will be available for antenna design, linkmargin prediction and evaluation of performance criterion of RoFSO systems in different deploy-ment environments.

ACKNOWLEDGMENT

Part of this work is supported by a grant from NiCT (Japan).

REFERENCES

1. Al-Raweshidy, H. and S. Komaki, eds., Radio over Fiber Technologies for Mobile Communica-tion Networks, First ed., Artech House Publishers, 2002.

2. Tsukamoto, K., K. Nakaduka, et al., “Development of DWDM radio on free space optic link sys-tem for ubiquitous wireless,” Asia-Pacific Microwave Photonics Conference (AP-MWP 2007),Apr. 2007.

3. Arimoto, Y., “Multi-gigabit free-space optical communication system with bidirectional beacontracking,” IEEJ Trans. FM, Vol. 127, No. 7, 385–390, 2007.

4. Marioni, F., Z. Sodnik, and F. E. Zocchi, “2.5-Gb/s free-space optics link over 1.1 km withdirect fiber coupling to commercial devices,” Free-space Laser Communications IV, Vol. 5550,60–69, Oct. 2004.

5. Nykolak, G., P. Szajowski, D. Romain, G. Tourgee, H. Presby, and J. Auborn, “Update on4 × 2.5Gb/s, 4.4 km free-space optical communications link: Availability and scintillationperformance,” Optical Wireless Communications II, Vol. 3850, 11–19, SPIE, Boston, Mas-sachusetts, Sept. 1999.

6. Kazaura, K., K. Omae, et al., “Performance evaluation of next generation free-space opticalcommunication system,” IEICE Transactions on Electronics Special Section on Evolution ofMicrowave and Millimeter-wave Photonics Technology, Vol. E90-C, No. 2, 381–388, Feb. 2007.

7. Aburakawa, Y. and T. Otsu, “Dense wavelength division multiplexed optical wireless linktowards terabit transmission,” Proc. MWP’03, 135–138, 2003.

8. Andrews, L. C. and R. L. Phillips, Laser Beam Propagation through Random Media, SPIEPress, Washington, 1998.

9. Andrews, L. C., R. L. Phillips, and C. Y. Hopen, Laser Beam Scintillation with Applications,SPIE Press, Bellingham, Washington, 2001.