![Schematic diagram of activated and inactivated nuclei condensation of continental aerosol [13].](https://www.researchgate.net/profile/Maha-Achour/publication/228702423/figure/fig3/AS:761728052768769@1558621528417/Schematic-diagram-of-activated-and-inactivated-nuclei-condensation-of-continental-aerosol_Q320.jpg)
![Attenuation due to (a) Selective and (b) Stable fogs [26].](https://www.researchgate.net/profile/Maha-Achour/publication/228702423/figure/fig4/AS:761728052756480@1558621528446/Attenuation-due-to-a-Selective-and-b-Stable-fogs-26_Q320.jpg)
Content uploaded by Maha Achour
Author content
All content in this area was uploaded by Maha Achour on May 23, 2019
Content may be subject to copyright.
UlmTech, all rights reserved.
- 1 -
Free-Space Optics Wavelength Selection: 10 µ Versus
Shorter Wavelengths
Maha Achour, Ph.D.∗
UlmTech, Inc.
731 S. Hwy 101, Suite 1E,
Solana Beach, CA 92075
Abstract
This paper is written in response to many inquiries about 10 µ Free-Space Optics (FSO) performance compared to
shorter wavelengths. The increasing interest in better understanding FSO weather effects is the result of carrier
requests as well as recent progress in analyzing fog effects on FSO signal propagation. Extensive studies in
modeling fog and simulating FSO attenuations revealed the complexity behind estimating FSO link availability in a
given geographical location. There are many different types of fog that are inhomogeneous along the propagation
path. Each type is characterized by the water droplet sizes and their concentrations, which are used in Mie scattering
theory to compute FSO signal attenuation. As a result, some vendors are augmenting their FSO links with a
microwave backup link or simply investigating other wavelengths claimed to be more resistant to fog such as 10 µ.
In this paper, we analyze ways to improve FSO link availability, 10 µ improvement compared to shorter
wavelengths, and challenges behind successful microwave backup installation.
Keywords: Free-space optics, optical wireless, atmospheric modeling, infrared propagation, 10 micron, 60 GHz,
Mie scattering, visibility, fog, link availability.
Introduction
When comparing FSO market research published in early 2000 with recent market research [4], we notice that the
more recent publications indicate a much clearer understanding of FSO technology and emphasize less hype about
its capability. The direct competitions between FSO vendors have led to a wide variety of products supporting
different optical and microwave technologies. FSO users can leverage these FSO improvements by selecting the
product suitable for their application. Readers interested in the FSO industry may wish to frequently check the FSO
Alliance progress and publications [12].
The current FSO market addresses the need for quick enterprise connections, cellular backhaul links, fiber cable
backups, and carrier metro/access deployment. Since the enterprise market, which is sometimes referred to by its use
of a campus-like installation, was the initial market deploying FSO, a clear understanding of link availability was
not required. The situation may have slightly changed with emerging enterprise real-time applications demanding
more reliable networks than previously. However, the enterprise market still does not need requirements that are as
stringent as those in the carrier market. It is for this simple reason that FSO will still be a player in the enterprise
market.
In the case of the fiber cable backup application, the point-to-point connection needs to comply with the FSO line-
of-sight requirement and limited distances. These types of installations usually do not require high link availability
due to the rare circumstances of fiber breaking in dense fog weather. For instance, when a few service providers
suffered network and phone service failures in New York, Free-Space Optics were quickly deployed because FSO
does not require licenses and is immune to microwave/RF interference.
∗ Contact email: machour@ulmtech.com; Web: http://www.ulmtech.com; Phone: 858-794 5322; Fax: 858-794 9628
UlmTech, all rights reserved.
- 2 -
The concept of extending phone lines through FSO has been around since the beginning of FSO deployment. The
need to interconnect cellular base stations with FSO is increasing due to 3G deployment and cellular phone
penetration. In some developing countries, cellular-phone services exceed wired-based telephone services. In
general, developing countries are very flexible with link availability because it is their only convenient
interconnectivity choice. This convenience is primarily due to FSO being inherently license-free nearly every where
in the world.. However, there have been exceptions, recently, in some developing countries when FSO deployment
monopoly is given to government-backed agencies. It is interesting to note that in some developing countries fiber
installation is sometimes more economical than FSO deployment due to low-cost manpower and no right-of-way
restrictions.
The carrier market is requesting high link availability in the 99.9%-99.999% range as well as a clear understanding
of FSO capabilities. This is due to the penalties carriers have to face when part of their network goes down. Network
operators control and monitor their network under the “no-excuse” rule: failure cannot be justified by the link
nature, location, or weather conditions. As we follow service provider business evolution, we notice that they are
gradually adding QoS-based services to respond to customer’s demands increase revenue and augment their ’dumb
data’ services. These added-value services are based on Service Level Agreements (SLA) between the provider and
end-users. As an example, Gold could be the service having the minimum packet drop, while silver, bronze and
best-effort being the alternative choices. In this case, service providers may certify FSO as the technology suitable
for some SLAs, and therefore customers serviced with FSO cannot get SLAs associated with the best QoS. This is a
shift from the famous 99.999% (five nines) availability, which allows a maximum of five minutes of failure per
year.
As we have mentioned earlier, the FSO spectrum is, in general, globally open without requiring deployment
licenses. This can be compared to unlicensed RF radios. License exempt bands, such as 0.9GHz/2.4 GHz/5
GHz/24 GHz/60 GHz, are open in a few countries. When addressing any unlicensed technology, immunity to
interference becomes the number one obstacle in dense deployment environments. Besides the 60 GHz case,
where Oxygen absorption eliminates interferences, most of the lower frequencies suffer from interferences when
adjacently deployed. Spread spectrum technologies are sometimes used to overcome interference.
Another advantage of FSO, when compared to RF, is the significant reduction in end-to-end delay. Again, with the
60 GHZ and higher frequencies exceptions, almost all radios have to process the radio signal to overcome channel
noise and interference while transmitting high throughput. This is achieved by using the so-called error correction
codes and high modulation levels, which are the two major elements in an RF system design that consumes the
highest delay. Long delay is not practical in a multi-hops deployment because of its accumulation in both transmit
and receive directions.
Most FSO products are plug-and-play units independent of the transmitted protocol and data rate. This feature is
vital to network operators since nodes are constantly upgraded to support new protocols and faster data rates. For
example, an OC3 FSO supports most of the protocols running up to 155 Mbps. Microwave, except for 60 GHz and
higher frequency radios, are limited to the transmitted protocol and speed of the signal being processed.
In order to boost network operators’ confidence in FSO compliance with their SLA offerings, FSO vendors need
to honestly present their FSO products’ performance in bad weather conditions to address users concerns and
establish realistic customer expectations. To date, two approaches have been offered to resolve the weather
problem: (1) augment FSO with an RF backup or (2) conduct thorough studies on weather conditions at each
installation. The first solution requires careful addition of the backup link, whereas the second choice calls for
extensive weather database and proper understanding of fog types in the region. In this paper, we address a
possible third choice that may eliminate the need for a backup link and exhibit improved fog penetration. It is
based on selecting 10 µ as the carrier wavelength over the current shorter 0.8 µ, 1.5 µ, 3 µ, and 5 µ wavelengths.
This paper is divided into four sections. In section one we review the different FSO technologies, commodity
components, and historic deployments. Section two will present the atmospheric effects and reviewing fog models
and their effects on the wavelengths of 0.8 µ, 1.5 µ, 3 µ, 5 µ and 10 µ. In section three we briefly review 10µ
technology and its FSO application. In section four we will compare fog attenuation of the above wavelengths based
UlmTech, all rights reserved.
- 3 -
on UlmTech’s FSO atmospheric modeling software, Simulight™ and previously published experimental and
theoretical results.
1. FSO Technologies
Free-Space Optical links consist of two transceivers communicating by means of emitting and receiving optical
signals. It is more practical to characterize the spectrum by its wavelength instead of its terahertz-frequency
spectrum. Figure 1 illustrates the electromagnetic spectrum and corresponding applications. Unlicensed radio
frequencies fall into two regions: the low (0.9, 2.4, 5, 24 GHz) and the high (60 GHz and higher) frequencies.
1.1 Unlicensed Microwave
In order to give the reader a clear comprehension of the benefits and limitations of each unlicensed band, we
summarized their properties in Table 1. As indicated, there have been many standards and activities in the
frequencies below 10 GHz due to their potential Local Area Network (LAN) interconnections. This has led to a
continuous drop in radio costs, but at the same time increased the interference risk. In Figure 2, we show the main
building blocks of radio and free-space optical systems.
Figure 2 shows the radical differences between a typical radio design and an FSO design. The main reason behind
the complexity of a radio design is the desire to transmit more bits in a given allocation of spectrum. In order to
accomplish this goal, many bits are carried within a single pulse. Information is carried within the pulse’s amplitude
(amplitude modulation) or phase (phase modulation). Since the loss of a signal translates into missing bits, error
corrections techniques are added for bit transmission redundancy [5]. IEEE 802.11 standards define the modulation
and error correction for the popular 2.4 GHz and 5 GHz spectrum. In addition to the building blocks illustrated in
Figure (2b), spread spectrum methods are imposed in these standards to combat interference when multiple radios
operate within the same region. These spreading techniques increase the transmitter and receiver signal processing
increasing the end-to-end communication delay. Standards-based radios are attractive because of their high quality
and lower production cost. Their main disadvantage is their low bandwidth and installation in an open environment
vulnerable to interferences.
The 60 GHz band spans 7 GHz of spectrum, and therefore their radios do not need to squeeze bits in the signal but
rather send one bit per pulse as in FSO. This makes 60 GHz end-to-end delay almost zero. Another feature that these
radios enjoy is being immune to interference. This is due to the Oxygen absorbing the weaker signal (side lobes)
around the main signal (principal lobe). However, the disadvantage of Oxygen absorption is that it further limits
deployment distances to few hundred meters to one kilometer. This is, of course, in addition to rain fade and other
applications constraints.
300MHz 30 GHz 300GHz 30THz 300THz
1m 10cm 1cm 1mm 100µ 10µ 5µ 1.5µ 0.85µ 100 nm 10nm
Frequency
Wavelengths
Radio Infrared Visible UV
1m 10-1m 10-2 m 10-3m 10-4m 10-5m 10-6m 10-7m
Near InfraredThermal Infrared Far Infrared
Millimeter Wave Microwave
0.9 - 24 GHz 60 - 120 GHz
Fi
g
ure 1: The electroma
g
netic s
p
ectrum and corres
p
ondin
g
a
pp
lica
t
ions
UlmTech, all rights reserved.
- 4 -
In general, radios operating above 10 GHz are affected by rain fade, and therefore their link availabilities depend on
the geographical location. For example, CLECS have installed LMDS radios, which operate in the 20-40 GHz
bands, 200 meters apart in Florida to guarantee 99.999% link availability. Recently, there have been activities
regarding the release of 70-90 GHz bands [6]. Unfortunately, the process will take time to overcome many
challenges related to these bands.
Table 1: Properties of unlicensed wireless spectrum.
Spectrum Band Technology Regulations & Standards
2.4 – 2.4825 GHz Worldwide Coverage, indoor/outdoor, 11 and 54 Mbps, up to 15
miles, DS spreading and OFDM.
ISM band: FCC part 15.247 and
15.249, IEEE standard 802.11b,g
B1 5.15-5.25
B2 5.25-5.35
B3 5.725-5.825
Limited global coverage, Hiperlan in Europe (B1 and B2),
indoor/outdoor (B2 and B3), 54 Mbps using OFDM on twelve non-
overlapping 20 MHz bands, and some non-IEEE radios support 450
Mbps using QAM.
UNII band: FCC part 15.407, IEEE
standard 802.11a
5.725-5.85 GHz Open in Asia and part of Europe, DS spread spectrum, and some
radios with 25 Mbps speed.
ISM Band: FCC part 15.247 and
15.249. No IEEE standard
24.05-24.250 GHz Radios in this band are provided by Sierra Digital and are affected
by rain.
FCC part 15.249. No IEEE
standards
57-64 GHz and
higher
Radios with Gigabit speeds, OOK modulation, no delay, few
hundred meters in range due to Oxygen absorption.
FCC part 15.255 and 15.249. No
IEEE standards.
200-300 THz Free-Space Optics (FSO) using short (785-850 nm) and mid (1550
nm) and long wavelengths (10 µ) signals, OOK modulation, no
delay, speeds up to 2.5 Gbps and few kilometers in distances.
Eye-safety IEC, FDA and ANSI
regulation, no FCC regulation, no
IEEE standards.
Conversion,
Amplification & OKK
modulation
Detection, Amplification
& Conversion
(a) FSO system
(b) Radio systems (excluding 60+ GHz)
Fi
g
ure 2: Hi
g
h la
y
er view of FSO and Radio S
y
stems
A/D and
Scrambler
Error
Correction
Modulation &
Pulse Shaping
Amplification & IF/RF
up-conversion
Detection and RF-
Baseband conversion
Channel
Channel
Demodulation
& Decoding
De-scrambler
D/A &
Amplification
UlmTech, all rights reserved.
- 5 -
1.2 Eye Safety
Eye-safe transmitted optical power standards have been created by the International Electrotechnical Commission
(IEC)[9], the Center for Devices and Radiological Health (CDRH) (which is part of the Food and Drug
Administration (FDA)), and the American National Standards Institutes (ANSI)[7]. The IEC60825-1 is in the
process of being amended, and the CDRH adopted the amended standard in May 2001, so there will be a single
worldwide standard for FSO devices. CDRH is a product safety standard, IEC combines product safety standards
and user’s safety guide, and ANSI Z136.1 is primarily a user safety standard.
In section eight of the IEC60825-1, FSO products are classified based on the energy carried within the signal pulse,
wavelength, and given transmission time period that covers consecutive pulse durations (long runs of “1” bits).
Class 1 products are considered harmless, whereas Class 1M is considered hazardous when viewed with optical
aided instruments. Since the outer layer of the eye, the cornea, acts as an optical band-pass filter passing
wavelengths ranging from .4 to 1.4 microns, light energy at wavelengths outside this range is absorbed by the cornea
and does not pass through the retina. It is for this reason that 1.5 µ and longer wavelengths are allowed to carry
higher power than the shorter wavelengths. IEC sets further restrictions on transmitted power to protect the cornea
itself from extended laser exposure. This is referred to as the Maximum Permissible Exposure (MPE). For example,
basic calculations of the IEC standard indicate that the MPE is around 1 to 2 mW/cm2 at 0.8 µ wavelengths and 100
mW/cm2 at 1.5 µ and 10µ. Some FSO vendors provide Class 1 products in the 1.5 µ and 10µ wavelengths.
1.3 FSO Systems
Free-Space Optical products are deployed in a line-of-sight point-to-point configuration. Each end consists of a
transmitter and receiver operating in a full duplex mode. That is, bits flow in both directions at the same time. FSO
systems can be installed on rooftops or behind windows, provided that glass attenuation is in the acceptable range.
In its simplest form, an FSO transceiver consists of an electrical sub-system and an optical sub-system. The
electrical sub-system amplifies the transmitted and received signals, modulates the transmitted signal, and monitors
and controls the unit’s operation. Most of the current FSO products support SNMP for remote monitoring and
control. The all-optical part interconnects lenses to fiber, sources and detectors. Information is transmitted between
the two transceivers using the On-and-Off Keying (OOK) method, which is a basic modulation scheme that
transmits a signal when the bit “1” is transmitted and transmits nothing for “0” bits. Long runs of consecutive zeros
and ones are important to understand when supporting different protocols not only to comply with eye-safety
regulations but also with the system optical requirements.
On the transmitter side, an LED, laser diode or fiber pigtailed laser is defocused slightly from the lens’s focal point
to incorporate beam divergence. The modulated signal is collected by the receiving lens and focused on to a
photodetector or fiber. Depending on the received signal intensity, “0” and “1” bits are recovered. Besides
photodetectors (quantum detectors), thermal detectors are common in the 3-5 µ and 8-14 µ wavelengths. Thermal
detectors respond to temperature changes from the absorbed optical energy.
In the presence of severe channel noise; the receiver introduces errors when “1” is decoded instead of “0” and vice
versa. With OOK modulation, direct detection (non-coherent) at the receiver end is used to demodulate the received
signal. In coherent detection, the signal shape is detected to derive both amplitude and phase information. The signal
carrier wavelength passes through an optical bandpass filter, which serves to reject background radiation which can
impinge on the surface of the detector. At the transmitter, lasers are commonly used to launch the optical signal, and
LEDs are sometimes used to reduce the cost of short-range FSO units. Since 0.8 µ wavelength transmitted power is
limited by the eye-safety standard (see section 1.2), off-the-shelf low power lasers fulfill most of the FSO
requirements. For 1.5 µ, however, more power can be transmitted and therefore vendors use high-power lasers or
Erbium Doped Fiber Amplifiers (EDFAs) to amplify the signal.
The photodetector operates within a certain power range of received light. That is, when the received light drops
below a certain minimum power, the receiver fails to properly operate because it is not receiving the minimum
UlmTech, all rights reserved.
- 6 -
number of photons to convert them into electrons. Commonly, this situation is referred to as receiver starvation. On
the other hand, when the photodiode receives too much light above its maximum threshold it will get confused and
fails to properly operate as well. This situation is referred to as receiver saturation. Since this paper will focus on
atmospheric signal attenuations, we will give two other examples of channel noise. The first one is background noise
when the receiver is placed in East-West configuration. Every year, and during a short period of time, the receiver is
saturated with sunlight. This automatically confuses the receiver by either saturating the detector or simply decoding
all transmitted bits as “1”. By observing the sun spectrum, background noise at short wavelengths is much higher
than the longer 10 µ wavelength. An example of system starvation is scintillation which will divert part of the beam
away from the receiving lens as discussed in section 1.4.
On the 10 µ technology side, Maxima Corporation is currently the only provider at this wavelength. Their products
are based on small, long life CO2 lasers, as well as, Quantum Cascade Laser which is a new class of semiconductor
laser emitting signals in the 8-12 µ atmospheric windows. The company uses its proprietary high-speed, room
temperature HgCdZnTe (MCZT) detector in the receiver. Besides applications in the communication industry, long-
wavelength technologies are strongly emerging in chemical, biochemical, biological and biomedical sensing,
imaging systems illuminators, infrared counter measures, vehicle collision warning systems (adaptive cruise
control), vehicle emission and combustion diagnostics, and military applications. As fiber-based communications
drove the current FSO components market, it is believed that military and security applications will drive the longer-
wavelength component market.
1.4 Scintillation
Atmospheric optical turbulence is due to the fluctuation in the index of refraction of the air. It affects the optical
signals temporal intensity and is referred to as scintillation. The twinkling of stars is a classic example of such an
effect. In addition to scintillation, beam wander and broadening also occur when optical signals pass through
turbulent air. FSO beams are subject to spatial and temporal fluctuations when passing through turbulence cells,
which have been extensively statistically modeled using weather parameters, altitudes and refraction index [3][10].
Small temperature variations cause virtual atmospheric lenses that can redirect the incident beam by refraction
effects. The result could blur the received beam spot (scintillation) reducing the received power, image spot dancing
at the receiver focal point, or completely divert the beam direction (beam wander).
Turbulence is greater closer to the ground or large horizontal planes such as roofs. It is for this reason that most, if
not all, FSO rooftop installations are deployed at the edge of the roof or use tall tripods to avoid heated air induced
by the roof itself. In order to study optical signal propagation through turbulence, some approximations are made to
solve the electromagnetic wave propagation equation. The main approximation is that the index of refraction does
not change rapidly in either space or time. One of the most widely used models is Rytov Models. It provides signal
power and phase fluctuations due to scintillations. For optical paths parallel to the ground, the equation is simplified
and shows a (2π/λ)7/6 dependency, where λ is the transmitted wavelength. From the inverse wavelength dependence,
it is clear that longer wavelengths are less susceptible to scintillation effects.
Aperture averaging over the receiver collecting area has been known to decrease scintillation effects. That means the
larger the receiving aperture is, the more scintillation it can combat. However, increasing the size of the receiving
lens is not practical, and therefore some vendors leverage spatial diversity by using several smaller apertures that are
sufficiently far apart that each received signal experiences independent air paths and therefore uncorrelated phase
and intensity variations. The summed output from such an array of detectors provides aperture averaging. A patent
has been issued describing the use of multiple apertures in FSO systems to combat scintillation [11] and reference
[10] analyzes the effectiveness of multiple apertures versus one aperture of identical area coverage. It was found
that multiple apertures tend to exhibit better averaging results for long ranges. Furthermore, the longer the
wavelength the less aperture averaging effects occur. This may be correlated to the fact that longer wavelengths are
less susceptible to scintillation in the first place.
Beam spot image dancing at the receiver plane is described by the angle of arrival. The variance of the angle-of-
arrival is found to be independent of the optical wavelength. Under large-scale atmospheric turbulence
inhomogeneities, the optical beam will experience random deflections as it propagates. This is referred to as beam
UlmTech, all rights reserved.
- 7 -
wander and is found to be independent of the wavelength. Experimentally, it has been shown that fast tracking can
correct the beam wander effect.
2. Atmospheric Effects
The atmosphere is a gaseous envelope surrounding the Earth. Due to gravitational force, which contributes to
atmospheric pressure, the maximum density occurs just above the surface and gradually becomes thinner as we
move away form the surface. In reference [2], the author analyzed the thermodynamics and statistics of the
atmosphere in detail.
Water vapor concentration is highest near the Earth’s surface; an indication that the Earth’s surface is the principal
source of atmospheric water vapor. Thermodynamically, water is the only substance in the atmosphere that occurs
naturally in all three phases with condensed water consisting of suspended water droplet particles (fog, cloud) and
hydrometeors (raindrops, snow). In addition, the atmosphere composition includes uncondensated aerosol particles:
dust, sea-salt particles, soil particles, volcano debris, pollen, by-products of combustion and other small particles
that arise from chemical reaction with the atmosphere. In general, aerosol particle sizes range between 0.1 and 10 µ
with drop-size concentration decreasing sharply with increasing drop sizes. Density refers to the mass of the
atmosphere per unit volume.
In the atmosphere we have a mixture of dry air and water vapor. In contrast to gases, which expand to fill the
volume of their container, condensed phases can maintain a free boundary (surface) and occupy a definite volume.
During droplet’s initial formations, the interface between the liquid and its vapor is not a two-dimensional surface,
but rather a zone that is several molecules thick, over which the concentration of water varies continuously until the
equilibrium state is reached. Water molecules at the surface of a droplet are subject to external interaction as well as
internal interaction with neighboring molecules but cannot freely move inside the droplet. When the forces
surrounding the surface molecules are asymmetric with inward net force, surface molecules will move towards the
interior of the droplet until equilibrium is attained. This effect is due to droplet surface expansion because the
energy of the surface molecule of the liquid becomes higher than the energy in the interior. Surface effect is due to
the surface tension work, which is proportional to the surface area, and the energy necessary to extend the liquid
surface against the vapor surrounding it. Therefore, the liquid surface is created as a result of an increase in the
potential energy. Spherical shapes have the smallest surface area for a given volume, and hence the smallest surface
energy. Hence, the spherical shape of cloud and fog drops resulting from the minimum-energy principle.
A fraction of the collisions between water vapor molecules in the atmospheric are inelastic, i.e. merging molecules,
leading to the formation of molecule aggregations which have a short lifetime since they disintegrate under
continuous molecule bombardment. When the aggregated water droplet reaches a size sufficient for survival, the
nucleation of water droplet then occurs.
Fog and cloud drops are formed by heterogeneous nucleation of some aerosol particles that are hygroscopic, i.e.,
attract water vapor molecules to their surface through chemical processes or physical forces caused by the presence
of hydrogen dipole. This effect may occur below 100% humidity. For example the deliquescent point of salt (NaCl)
is about RH=75%. Another factor contributing to droplet growth is the collection model where small drops collide to
form larger drops. One efficient model of such a phenomenon is called the stochastic collection model that accounts
for the probabilistic aspects of collision and coalescence.
Cloud condensation nuclei (CCN) are considered a subset of hygroscopic aerosol particles where nucleation of pure
water drops may occur at supersaturating less than 1%. For drops below the critical nucleation radius r*, the drops
grow only in response to increase in relative humidity and are termed haze particles. A condensation nucleus is said
to be activated when the drop formed on it grows to size r*. Once the drop grows slightly beyond r*, its equilibrium
value of supersaturation is less than the supersaturation at the critical radius formation. As it will be explained later,
droplet size distribution in the atmosphere will sometime exhibits two radius peaks instead of one. This is referred to
as bimodal distributions. Stochastic collection, along with the equilibrium of droplets with its surroundings,
contributes to the bimodal size distribution explanation.
UlmTech, all rights reserved.
- 8 -
The explanation of the observed broadening of drop size spectra n(r) remains a major challenge to cloud and fog
physicists. Fog is composed of very small water drops in suspension in the atmosphere which greatly reduces
visibility. It is mainly caused by cooling of air to its dew point and the addition of water vapor in the air. Fog may be
considered as a stratus cloud whose base is low enough to reach the ground. Fog, stratus, and stratocumulus clouds
occur within 2 km from the ground. They are not associated with precipitation, although often produce drizzle.
2.1 Aerosols
The atmospheric aerosol forms a particle spectrum which spans over four orders of magnitude sizes, and are
subdivided into four categories as described below and illustrated in Figure 3a [13]:
- Aitken particles: The smallest particles are less than or equal to 0.1 µ.
- Larger particles: Their corresponding sizes ranges between 0.1 and 1 µ.
- Giant particles: For radius greater than 1 µ.
Maritime Fog
0.01
0.1
1
10
100
1000
0.02 0.064 0.2 0.64 2 6.4 8 >8
Droplet Diameter in Micron
Concentration
Continental Fog
0.01
0.1
1
10
100
1000
10000
0.02 0.064 0.2 0.64 2 6.4 8 >8
Droplet Diameter in Micron
Concent ration
It is important to emphasize that Aerosol terminology in the context of our studies includes nuclei resulting from
water vapor condensation regardless of their physical or chemical properties. In Figure 3a, large particles are
responsible for scattering visible and near infrared rays. Furthermore, only large and giant nuclei become activated
as condensation nuclei because the required water vapor supersaturation is always small. Aitken nuclei, however,
remain inactive with a smooth line separating the active from the inactive nuclei [13] because of the higher
supersaturation levels required to condense Aitken particles. This is another phenomenon contributing to the
Air Electricity
Air Chemistry
Cloud Phys.
Atm Optics
Size
Ranges
for:
10-3 10-2 10-1 100 101 103
Particle radius in
µ
Large
Small
ions
Classification
Active CCN
Haze
Large ions
Aitken Giant
Particles mainly contributing
to Aerosol Mass
Fig. 3a: Classification of Aerosol Particles [13].
Fig. 3b: Example of particle concentrations for maritime and continental regions
UlmTech, all rights reserved.
- 9 -
bimodal water droplet distribution models. As relative humidity increases, Aerosol particles gradually grow to
become cloud, fog or ice droplets.
In general, many methods of measuring particle size distribution are simultaneously needed to cover the whole
spectrum of Figure 3a. Due to this complex measuring mechanism, frequently one method is used restricting the
outcome to a subset of the whole spectrum [13]. For instance, particles below 0.1µ are commonly derived from the
diffusion parameter whereas particles between 0.1-0.5 µ may be collected using electrical or thermal precipitation.
Particles larger than 0.5 µ are derived from light scattering instruments. Furthermore, it has been noted that Aerosol
size concentration is further classified between continental and maritime measurements as indicated in Figure 3b
[13].
The most important properties of an Aerosol mix are their size variations and relative humidity. Particle radius
grows until equilibrium between the droplet and its surroundings is attained. For soluble particles, equilibrium
depends on the initial particle size and its salt mass content. At equilibrium, the relative humidity of salt droplets
indicates that the influence of the dissolved salt decreases more rapidly with increasing size than does the influence
of surface curvature. Therefore, the growth curve passes the 100% humidity line, reaches a maximum and then
approaches the 100% humidity line asymptotically. In other word, water vapor condensation requires the particle to
be in an unstable supersaturation state with its surrounding to continuously grow. For giant and large particles, the
critical supersaturation amounts to only few tenths of a percent (for example 100.01%), whereas it increases rapidly
to several percents (for example 100.1%) for Aitken particles. As mentioned earlier, this phenomenon also
contributes to the bimodal size distribution explanation.
The supersaturation is of great importance for cloud and fog formation. When air is cooled and fog droplets start to
form, only the largest condensation particles with lowest supersaturation start to grow. Smaller particles, however,
often do not reach their peak because they require much higher supersaturation level and remain on the stable branch
of the growth curve. We schematically illustrate this effect in Figure 4 to explain the bimodal size distribution effect.
Radius
10-2 100101
102
103
RH = 70%
inactivated
100%
Activated
100%
Combined
100%
dN/d log(r)
per µ cm-3
Fig. 4: Schematic diagram of activated and inactivated nuclei condensation of continental
aerosol
[
13
]
.
UlmTech, all rights reserved.
- 10 -
At higher altitudes, supersaturation reaches higher values making the Aitken and large particles activated. Therefore,
Figure 4 exhibits one peak at much higher radius sizes at these altitudes. It is for this reason mid and high clouds
droplet distribution shows one narrow peak concentration at giant particle sizes.
In most literature, the basic assumption in size distribution calculations is that the temperature and humidity fields
are completely uniform throughout the fog or cloud. In reality, there are always fluctuations that are depicted in the
gap between the inactivated and activated curves in Figure 4. Due to aerosol combination of mixed nuclei and the
non-linear dependence of supersaturation with the particle size, the dividing gap between the inactivated and
activated particles is smoothed out giving birth to a second maximum, i.e. bimodal distribution size distribution
function.
2.2 Attenuation
The atmospheric transmission of optical signals, τa, is described by the following Beer’s law equation:
Where, βabs and βscat are the absorption and scattering coefficients, respectively. In some literature, τa is also referred
to as the atmospheric transmittance or simply the transmittance of optical depth R. Absorption by atmospheric gases,
such as H2O and CO2, defines the atmospheric windows where absorption is less severe. Scattering, on the other
hand, is due to small particulates in the transmission path of electromagnetic waves, abstracting their energy and re-
radiating it in different directions. More detailed analysis and information about absorption and scattering could be
found in the author’s SPIE publications [1][2].
Since visibility is the only weather parameter that describe different fog situations, scientists have tried to directly
relate the scattering coefficient βscat to the incident wavelength λ and visibility. Most of the early FSO publications
used the famous Kruse relation [14]:
Where, the exponent is δ = 1.6 for good visibility, 1.3 for V=6-50 Km and 0.585 V1/3 for visibility less than 6 Km.
The problem with equation (2) is that it fails to estimate attenuations for visibilities below 1.5 Km [2]. Therefore, the
author relied on the original Mie equation to estimate attenuations due to fog and rain [1][2]. For the Aerosol
distribution discussed in section 2.1, the total scattering coefficient is derived from the following equation:
Where, Qscatt is the scattering efficiency, r is the Aerosol particle radius, nr is the concentration of radius r particles
and λ is the transmitted signal wavelength. Therefore, different types of fog are described by their water droplet
concentration nr which reproduces the measured visibility V. From equation (2), βscat is related to visibility at λ= 550
nm (which is the most sensitive wavelength to the human eye) via the following equation:
The correlation between particle sizes and incident wavelength is given by the term “r2 Qscatt(x)” in equation (3). In
Figure 5 we have plotted “r2 Qscatt(x)” versus the droplet diameter “2r” for the wavelengths 0.5 µ, 1.5 µ, 3 µ, 5µ and
10µ. It is clear that for fog particles less than 5µ in diameter, 10µ light is the least affected and becomes more
affected by particles of diameters 10µ and above. The contribution of these giant Aerosol particles dominates
attenuation and visibility measurements.
(1) e R ) ( -
a
scatabs
ββ
τ
+
=
()
(3)
r 2
, )(β2
r
scatt x x Q nr scattr
λ
π
πλ
== ∑
(2)
55.0
3.91
55.0
)0.55(λ ) (βScatt
δδ λλ
µβλ
−−
=
== V
Scatt
(4)
)550(β
3.91
V
scatt
nm=
=
λ
UlmTech, all rights reserved.
- 11 -
From the above discussions we conclude that, in general, the 10 µ light suffers less attenuation than shorter
wavelength light. This is because the concentration of Aitkin and large particles are always larger, if not much
larger, than the giant droplets at low altitudes. In section 4 we will display some experimental as well as theoretical
results of wavelengths performance in different types of fogs.
(a) (b)
Figure 5: Attenuation versus droplet size diameter for different wavelengths
In summary, fog is the most challenging media for FSO signals to penetrate. Fog attenuation is wavelength
dependent and can severely attenuate the signal depending on the fog type. Most common fogs attenuate 10 µ
signals much less than shorter 0.8 µ - 1.5 µ wavelengths. Hence, 10 µ FSO systems are expected to perform better
than shorter wavelengths in fog. On the other hand, Rain attenuation is independent of the transmitted wavelengths.
3. “10 µ” Technology
For over 30 years, the DoD has pursued 10 µ in the battlefield because of its high penetration in different
atmospheric conditions, smoke, and dust and has deployed 10 µ CO2 lasers in spite of their previous limitations:
large size, high power consumption and inability to modulate at high speeds.
Besides applications in the communication industry, these long-wavelength technologies are strongly emerging in
the following industries:
- Military: Since 1970, the military has pursued 10 µ in the battlefield because of its high penetration in
different atmospheric conditions, smoke, and dust. The DoD has deployed 10 µ CO2 lasers in spite of their
previous limitations: large size, high power consumption and inability to modulate at high speeds. With
recent security focus, this technology may play a potential role in the chemical and biological sensors.
- Medical: Analysis of breath for the early detection of several potentially deadly diseases such as ulcer,
colon cancer and others. Its biomedical spectroscopy capability detects and analyzes common molecules of
blood plasma.
- Automotive: Vehicle collision warning system such as adaptive cruise control, monitoring and sensing
chemical and combustions in vehicle emission.
- Law enforcement: QCL-based sensors to detect explosives, emission from illicit drug production sites and
other hazardous chemical and biological substances.
UlmTech, all rights reserved.
- 12 -
- Others: In many scientific areas such as atmospheric studies where QCL is used to determine chemical
concentration profiles.
Due to these critical applications, there have been activities in building and commercializing 10 µ lasers more
practical than the CO2 laser. One of the leading emerging long wavelength lasers is Quantum Cascade Laser (QCL).
Maxima Corporation, the only FSO vendor working in the long wavelength range, has developed a QCL laser as
small as other semiconductor-based lasers with 130milliwatts of optical power in the 8-12 µ atmospheric window.
On the receiving end, Maxima has also developed its proprietary HOT (high temperature photodetector) high-speed
mercury-cadmium-telluride HgCdZnTe (MCZT) as seen in figure 6, which operate with sensitivities near
Background Limited Performance.
Fig. 6. Maxima 10micron integrated photodetector
Semiconductor quantum cascade lasers (QCLs) achieve high power emission at room temperature and operate at
various wavelengths between 5 µ to 17 µ. The thickness of the used material determines the wavelength of the
emitted laser signal. Each QCL is a collection of alternating InGaAs and InAlAs semiconductor layers on an InP
substrate. The corresponding electrons cascade down the path with a probability of emitting laser photons every time
they hit a new layer.. MAXIMA Corporation has recently demonstrated the first commercial 10µ FSO systems
utilizing these devices with clear air link margins (fade margin) similar to 1µ micron systems. Given that 10µ
wavelengths have significantly lower attenuation in fog (discussed in the next chapter) this link margin gives 10µ
technology several times distance advantage over 1µ systems.
Reference [23] provides a good summary of Quantum Cascade Lasers.
4. Wavelength Performance Comparison
We start our wavelength comparison section by reviewing the results of reference [26]. Arnulf et al. measured
attenuations of wavelengths ranging between 0.35µ to 10µ in haze and in fog. Haze conditions are characterized by
visibilities greater than 1 Km, where droplet sizes are less than 1 micron (Figure 3a). Measured optical transmission
through haze was found to increase remarkably with increasing wavelengths. Arnulf et al. also analyzed and
measured attenuations due to selective, evolving and stable fog. Figure 7 shows some of their results. By simply
observing these two figures, we realize the complexity of quantifying fog attenuations.
Selective fogs affect some wavelengths more than others as illustrated in Figure 7a [26]. One notices that
wavelength transmission depends on the corresponding visibility. For visibilities above 200 m, 10 µ attenuation is
found to be less than 10 dB/Km, whereas 1µ attenuations varies between 10 and 60 dB/Km. This is a clear
indication of giant fog droplets absence. In the case of visibilities less than 100 m, 10µ may be subject to attenuation
as high as 100 dB/Km which is still one tenth of 1µ attenuation. At these short visibilities, some giant particles start
to formulate.
In figure 7b we display attenuation versus wavelengths in Stable fog. Unlike selective fogs, where attenuation is
more sever for some wavelengths, Stable fogs tend to attenuate most wavelengths in a similar way with some
improvement in the 10µ range. It is clear that Stable fog contains more giant particles than selective fog and the
concentrator
absorber
heterojunction contacts
s ubstrate
reflector
metalliz ation
GaAs
HgCdZnTe
concentrator
absorber
heterojunction contacts
s ubstrate
reflector
metalliz ation
GaAs
HgCdZnTe
UlmTech, all rights reserved.
- 13 -
lifetime of Stable fog is short at low altitudes. In this case, 1µ attenuation is generally twice as much, on a dB/Km
log scale, as 10 µ attenuation.
Simulight™ [1][2][18] is software developed by UlmTech to model clear, rain, evolving, selective and stable fogs.
As discussed earlier, bimodal fog droplet distributions occur for very low visibilities conditions with the first
maximum occurring in the submicron to one micron region. This automatically indicates lower attenuations at the
10µ wavelength compared to shorter wavelengths. In figure 7b, we selected a few reference points to compare
Arnulf et al. Stable fog results with Simulight results, which are shown in the six black dots. We notice good
agreement between Simulight™ modeling and reference [26] measurement for Stable fogs.
Next we briefly cover backscattering effects. In section 2 we studied forward scattering, which refers to light
scattered by the droplet and aimed in the forward direction. Backscattering refers to scattered optical signals and
(
a
)
(
b
)
Figure 7: Attenuation due to (a) Selective and (b) Stable fogs [26].
Simulight
UlmTech, all rights reserved.
- 14 -
directed in the opposite direction of the incident light. Reference [17] incorporates backscattering calculations into
their theoretical attenuation models and compares them to experimental results. Wavelengths ranging from 0.7 µ to
10.6 µ are considered while focusing on the 10.6 µ improvement.
It was found that fogs with maximum droplet concentration around and above 10 µ start to affect 10.6 µ and all
other wavelengths as well. This is easily seen in figure (5b) where all lines are intersecting around that droplet sizes.
In addition to scattering coefficient, the paper also covers the backscattering coefficient and found that, unlike
forward scattering, backscattering is wavelength dependent for droplet diameters around and above 10 µ. Using two
fog models and comparing them to experimental data [17], transmission using 10.6 µ exhibited the lowest
backscattering coefficient compared to 0.7 µ, 1.06 µ, 2.8 µ, 3.5 µ, 8 µ and 5.3 µ, with the latter having the highest
backscattering value. Similar calculation and experiments were conducted [17] in rain and 10.6µ, again, showed the
lowest backscattering coefficient with 1.06 µ having highest value.
Incorporating backscattering effects is more complicated than forward scattering, and therefore it is often omitted
from most publications analyzing optical signal transmission in different weather conditions. It was also omitted in
Simulight™ [18] because it required a long run-time convergence.
In reference [19], fog droplet-size distribution was studied by measuring the transmission measurement of
wavelengths 0.53µ, 0.65µ, 1.23µ, 2.2µ, 4.01µ and 10.1µ over a 17 m path. Choosing such a short path is more
efficient for studying fog because it eliminates the path averaging of attenuation and hence droplet concentration.
Measurements were made on 29 of January 1981 and recorded four hours before fogs were formed until their
dissipation. The interesting result is that most of the fog droplet concentrations had two peaks instead of one peak
with the first peak occurring in the sub-micron region. This was in agreement with bimodal distributions reported
earlier by Eldrigde [20] and on which basis Simulight™ [18] was built. Reference [19] also concludes that fog
having concentration of droplets diameter in the 3.2-12.8µ are short-lived dense fogs.
Reference [22] provides a good summary of optical communications systems in the 1.06-3.0µ near-IR (infrared),
3.0-5.0µ mid-IR and 9-11µ far-IR wavelengths. Experimental results found that Far-IR coherent detection shows
superior performance in low visibilities. And in a most recent publication [24], a free-space optical link using 9.3µ
quantum cascade laser and an HgCdTe detector was performed over a 350m range. Under foggy conditions with
visibility below 100m, the received signal was subject to only 20%, or 21 dB/Km, attenuation in this dense fog. That
is 80% of the transmitted signal is recovered at the receiver.
On the experimental side, reference [25] conducted experiments using three types of spectral regions:
- Visible: 0.4µ to 0.7µ
- IR: 3-5µ and 8-12µ
- RF: 94 GHz, 32 GHz and 10 GHz
It was observed that attenuation due to moderate rainfall (up to 50 mm/hr) is around 14 dB/Km and was identical, as
expected, for visible and IR signals. For evolving fogs, attenuations decrease rapidly with increasing wavelength,
whereas for stable fog IR signals exhibit slight advantage over visible signals. Clouds seem to have drop-size
concentrations around and above the 10µ diameter, and therefore they affect visible and IR the same way.
As for Radio Frequency (RF) signals, it was found [25] that they are transparent to any type of fog conditions with
only 2 dB/Km (35 GHz) and 3-10 dB/Km (94 GHz) attenuations for 30-meter visibility condition. Similar results
are concluded in cloud measurements. Attenuation due to rainfall, however, is negligible for 10 GHz and could be
sever for 35 GHz and 94 GHz reaching 10 dB/Km and 30 dB/Km, respectively, for 100 mm/hr rainfall.
Conclusion
We conclude based on measured field data of short (0.8 µ), mid (1 µ) and long (10 µ) wavelength light, that long
wavelengths typically have better fog penetration as high as twice (stable fogs) and ten times (selective fogs) the
transmission of shorter wavelengths. This result cannot be generalized to all types of fogs because of the unknown
UlmTech, all rights reserved.
- 15 -
distribution of particle sizes. However, from the statistical and thermodynamics studies, it is conceivable to conclude
that long lifetime fogs primarily have distributions of small particle sizes.
In this paper we analyzed the challenges, benefits and performance of different wavelengths from the visible, 1.5µ,
3µ, 5µ and 10µ wavelengths. Besides propagation through fog and rain, we analyzed scintillation effects,
backscattering and compliance with eye-safety standards. It was found that 10 µ has a significant advantage over
shorter wavelengths for FSO communications systems due to its better performance in fog and that components now
exist with adequate performance levels to produce commercially viable 10µ FSO products.
The author would like to thank Maxima Corporation [27] for discussions and providing 10 µ component references.
References
1. Achour Maha, Simulating atmospheric free-space optical propagation: rainfall attenuation, Proceedings of
SPIE Vol. 4635 (2002).
2. Achour Maha, Simulating Atmospheric Free-Space Optical Propagation: Part II, Haze, Fog and Low
Clouds Attenuations, Proceedings of SPIE Vol. 4873 (2002).
3. The Infrared Handbook, Chapter 6, Wolfe and Zissis [1985].
4. Bettina Tratz-Ryan and John Mazur, Free Space Optics- Focus Report, Gartner Group, September 2002.
5. Digital Communications: Fundamentals and Applications, Bernard Sklar, Prentice-Hall publisher [1988].
6. [FCC] FCC Initiatives Proceeding to Promote Commercial Development of Spectrum in the 71-76 GHz,
81-86 GHz and 92-95 GHz Band, FCC [2002].
7. [ANSI] http://webstore.ansi.org/ansidocstore/default.asp
8. [FDA] http://www.fda.gov/cdrh/comp/rad_consumer.html
9. [IEC] http://www.iec.ch/
10. Laser Beam Scintillation with Applications, Andrews et al., SPIE Press 2001.
11. Multiple transmitter laser link, Eric Korevaar, US Patent 5,777,768 [1998].
12. Free-Space Optical Alliance, www.wcai.com/fsoalliance
13. C. E. Junge, Air Chemistry and Radioactivity, Academic Press, [1963].
14. P. Kruse et al., Elements of Infrared Technology, John Wiley & Sons, [1962].
15. Lasers and Electro-Optics: Fundamentals and Engineering, Christopher C. Davis, Cambridge University
Press, [1996].
16. Semiconductor-Laser Fundamentals, W. W. Chow, Spinger Publisher, 1999.
17. Comparative Studies of Extinction and Backscattering by Aerosols, Fog and Rain at 10.6µ and 0.63µ, D.
B. Rensch and R. K. Long, Applied Optics, Vol. 9 No. 7, pages 1563-1573 [1970].
18. Simuligh:FSO propagation in clear, fog and rain Software, UlmTech Inc. www.ulmtech.com/FSO.html
19. Drop-size distribution of fog droplets determined from transmission measurements in the 0.53-10.1 µ
wavelength range, A.P. Lenham and M. R. Clay, Appled Optics Vol. 21, No. 23, pp 4191 [1982].
20. Haze and Fog Aerosol Distributions, R. G. Eldridge, Journal of Atmospheric Sciences, Volume 23, page
605, September [1966].
21. J-5 Optical Heterodyne Communications Experiments at 10.6µ,F. E. Goodwin and T. A. Nussmeier, IEEE
Journal of Quantum Electronics, pp 612, October 1968.
22. Imaging Laser Radar in the Near and far Infrared, G. R. Osche and D. S. Young, Proceedings of the IEEE
Vol. 84, No. 2, pp103 [1996].
23. Quantum Cascade Lasers, Capasso et al., Physics Today, May 2002.
24. Free-Space Optical Data Link Using Peltier-Cooled Quantum Cascade Laser, Blaser et al., Electronics
Letters, Vol. 37, No. 12 [June 2001].
25. Attenuation of Electromagnetic Radiation by Haze, Fog, Clouds and Rain, C. C. Chen, RAND Report
[1975].
26. A. Arnulf et. al., Transmission by Haze and Fog in the Spectral Region 0.35 to 10 Microns, Journal of the
Optical Society of America, volume 47, number 6, [1957].
27. [Maxima Corporation] http://www.maximacorp.com
