Chapter 1

Introduction to Optical
Communication Systems

The purpose of this chapter is to give an overview and a somewhat historical perspective on the eld of
optical communications. The rst section of the chapter describes why there are fundamental reasons that
optics is attractive for use in communications. Indeed, the worldwide telephone network, the worlds largest
communications system, is ber optic-based. Further, optics is nding progressively more communications
applications by the day, not to speak of the numerous sensing applications that are presently being imple-
mented. In some areas, optics is being implemented somewhat more slowly than originally predicted. This
is in great part due to cost. In the telephone network, hardware cost was really not an issue compared to a
myriad of other costs such as right of way and cable installation, and the initial implementations were carried
rapidly rather than cost-eectively. The telecommunications solutions, which were the rst to be carried
out, could therefore not be taken over directly into other applications. Now, though, the telecommunications
market is competitive as are the data communications market and others, and various new cost-eective so-
lutions are appearing. There are numerous applications where optics is competing. The second section of
this chapter discusses a very general systems model which can be used to describe essentially any optical
system in use at present. The third, fourth, and fth sections give a somewhat historical perspective on
the development of the three most important system components, the source, the ber, and the detector.
Section 6 gives a more broad systems perspective.

1.1 The Electromagnetic Spectrum and What It Tells Us about

the Relative Advantages of Optical Communications
Various observations relating to the potential of optical technology for transmission of information can be
made from perusal of the frequency line of Table 1.1. The information rates in which one is interested
in conventional modern-day communications systems generally correspond to audio rates (or multiples
thereof) in telephone systems, radio rates (or multiples thereof) in commercial broadcast systems, or digital
television rates (or multiples thereof) in the most advanced of video distribution systems. These rates (using
a digital television rate of 140 Mbit/sec times perhaps 30 channels) are generally below several gigahertz.
If one were to transmit such information without impressing it on an optical carrier but instead on a radio
frequency (RF) carrier a bit higher than the maximum rate, the transmission wavelength of the RF carrier
would be centimeters or larger. There can, however, be great advantages to using optical carriers. An obvious
one is the low loss and directionality of the optical ber. Clearly, the carrier must have a higher rate than
the information rate. A major principle that has, to present, appeared in communications systems has been
that the higher the frequency, the greater the technical complexity. Microwaves are harder to handle than
are radio waves. As wavelengths decrease to approach the size of circuit components, circuit elements are no
longer lumped as such, and leads can act as reective components and/or antennas and lumped elements as

1
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 2

300km 300m 30cm 3mm 30m .3m 30

A .03
A

name audio radio wave mmwave IR OPT/UV x-ray gamma

103 106 109 1011 1013 1015 1017 1020

KHz MHz GHz THz

p 3eV .3meV 30meV 3eV 300eV .3 MeV

Table 1.1: A frequency line which gives the wavelengths , the frequencies , and the photon energies h for
the various regions of the frequency spectrum named on the line.

electromagnetic resonators (as will be discussed further in Chapter 4). This has generally meant that sending
more information would cost more and that there was therefore, in some sense, a cost per bit/sec (bps) of
transmitted information in the sense that going to a higher information rate requires a higher frequency.
Thus, the rst observation from the frequency line would be that, for optical carriers, which have frequencies
in the hundreds of terahertz, information bandwidth is in some sense free. That is to say, the optical
wavelength is so small compared to most devices that the technology has changed drastically from electrical
and microwave. Once we assume that we have such technology, then no matter how high an information
rate one might want it will not be necessary to change the carrier, as the carrier frequency is higher than any
realistic information rate could become. Bandwidth is not completely free, though, as encoders and decoders
must necessarily operate at the information rate, but much of the rest of the system must necessarily handle
only the carrier plus modulation. If a component can handle a frequency of 5 1014 hertz, an information
shift in that frequency of a part in a thousand (corresponding to a 500 gigahertz information rate) will have
little or no eect on device performance. Therefore, once the system is already set up, one can upgrade
system speed more or less at will without the kind of costs incurred by changing the electromagnetic (i.e.
suboptical) carrier in conventional systems. (This is not quite true in long-haul communications at very high
bandwidths where ber dispersion can limit the repeaterless transmission span, but this is a rather special
case. However, in any of the non dispersion-limited links, active eorts to increase throughput as trac
increases are being employed. As will be mentioned below, important cases of these are (dense) wavelength-
division multiplexing (DWDMto be discussed in section 11.3 of Chapter 11) and code-division multiple
access (CDMAto be discussed in section 13.?? of Chapter 13).)

A consequence of the size of the optical bandwidth is that the optical carrier can be used to carry many
dierent telephone conversations, television programs, etc., simultaneously. The process by which this is
generally carried out (at least in synchronous format) is called time division multiplexing (TDM). The idea
is that, if one wishes to multiplex 16 dierent channels each transmitting at 1 Mbps, one could perform
this by dividing each bit period by 16 and then interleaving the bits into a composite 1 sec bit (1 Mbps
rate) which actually carries 16 bits of information on it. With telephone conversations representing a rate of
64 kbps, the 100s of Tbps bandwidth of the optical carrier holds great promise for TDM. Of course, TDM
is not the only multiplexing scheme one can imagine using. One could imagine impressing a number of
subcarriers, spaced by perhaps some gigahertz, onto the optical carrier. Each of these carriers could then be
modulated at an information rate and then reseparated according to their dierent carrier wavelengths at the
output. Such a scheme is referred to as wavelength-division multiplexing (WDM) or subcarrier modulation,
depending on the implementation. Many of the present-day schemes for increasing link throughput with
increasing trac involves combining many TDMd signals onto WDMd carriers. In fact, the limitation on
density of WDM turns out to be not bandwidth but power. That is, each channel requires some amount
of power. The more channels, then, the higher the power requirement. At some power level, optical ber
nonlinearity becomes important, and this nonlinearity tends to mix the signals together. There is presently
much eort going on in trying to nd ways to equalize such nonlinearities.
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 3

The high carrier frequency of the optical carrier also has some drawbacks, especially as it relates, through
the speed of light, to the optical wavelength (Table 1.1). The optical period corresponds to less than two
femtoseconds. This means that phase control corresponds to manipulation of subfemtosecond periods of
time. Although techniques to do such are emerging, they are complicatedmuch more complicated than
manipulating microwave or radio frequency waveforms. For this reason, coherent optical reception is still
a laboratory technology. The development of the rare earth-doped optical ber amplier seems, for the
present at any rate, to have obviated the need for coherent techniques in telecommunication as far as
improved signal-to-noise ratio goes (as will be discussed in section 11.4 of Chapter 11).
The short period of the optical wave also implies a short wavelength centered around a half of a micron.
The smallness of the optical wavelength, therefore, promises to allow for the miniaturization of transmit and
receive modules, which should allow considerable reduction in size, weight, and cost of optical communica-
tion systems with respect to microwave/radio wave counterparts. One needs to be careful about the size
scaling, however. Although the ratio of the optical to the microwave wavelength is roughly 105 , waveguide
dimensions do not quite scale with this factor. A fully metallic (closed) waveguide must always have a
dimension of roughly /2. This is not the case with fully dielectric waveguides such as the ones necessary
for optical bers or integrated optical devices. The smallness of the optical wavelength dictates the use of
dielectric guides, as even a tiny metallic loss per wavelength would incur a huge loss over propagation dis-
tances of millions to billions of wavelengths. In this dielectric waveguide case, the waveguide core dimension
is /(2NA), where NA is the numerical aperture. The numerical aperture is directly related to the index
contrast between the core and cladding of the dielectric waveguide. By nature, the index contrast must be
low, as if it were not, this would indicate that there was a big dierence between the core and cladding
material, which would naturally lead to a large scattering loss. A typical NA of 0.2 leads to waveguides with
a characteristic dimension of about 5, or a factor of 10 larger (per wavelength) than a metallic waveguide
counterpart if one could be made to propagate light. This factor of ten would mean that optical dielectric
waveguides could still be a factor of roughly 104 smaller than a microwave guide. Actually, the dierence
with electrical is yet smaller. Microstrip and coplanar waveguide circuits are open waveguides combining
metal and dielectric characteristics. The characteristic length of an electromagnetic wave is necessarily its
wavelength. The characteristic length of a current, however, is the electron wavelength, which can be in the
angstrom range. Current can be conned into tiny metallic strips much smaller than the electromagnetic
wavelength. As the microwave wavelength is so much longer than the optical wavelength, the metal loss is
not nearly so crucial as in the optical case. This is the situation in microstrip and coplanar waveguides,
where characteristic stripe dimensions can be tens of microns in extent and therefore comparable to ber
outer diameters. The current, however, only serves to guide the electromagnetic wave which actually carries
the signal. Although the most intense, near-eld portion of this wave can be conned between conductor
and ground plane, the wave clearly can sample its environment out to much greater distances. Therefore,
the higher the packaging density of such open microwave channels, the worse the crosstalk. No matter how
tightly one packs ber, the crosstalk is essentially zero if the cladding is properly designed. This leads to the
characteristic that ber is an excellent medium for space division multiplexing (SDM)that is, packaging a
number of channels with dierent information streams in close proximity.
Although all the advantages of coherent optical communication systems have yet to be brought to fruition,
another property of optical radiation has made todays optical communication systems not desirable for
applications. The important property here is that of photon energy. As is seen from Table 1.1, the photon
energy ranges from roughly 2 eV to roughly 4 eV. As one perhaps recalls from freshman physics, the hydrogen
atom, that atom with one of the most tightly bound unpaired outer valence electrons (helium paired outer
electrons will have a higher ionization potential), has an ionization energy of 1 rydberg, which is 13.6 eV.
As other atomic and molecular transitions must therefore correspond to a fraction of a rydberg, this means
that photons can be used for photoionization as well as pumping of atomic transitions. This leads to signal
dispersion, as a nite signal linewidth (information impressed on the carrier) will cause one side of the line
to lie closer to one transition than to the other, and therefore the two ends of the line will see slightly
dierent media. As the room-temperature phonon energy is 26 meV, single optical photons are detectable
with solid-state detectors. Microwave signals are not and require (inecient by minimally 3 dB) antennas.
This would seem to be an advantage in eciency. However, there is also a penalty to be paid for having
such a photon energy. Because single photons are detectable, the emission/reception process must take on
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 4

a granular nature. As is well-known, even in a steady rain, the probability of a raindrop landing (as a
function of time) follows a Poisson distribution, implying that there is raindrop bunching. A raindrop would
rather fall right after the one before. Raindrops are impatient and dont like to wait. In much the same
manner, a laser likes to spit bunches of photons even under constant bias current. Such behavior leads to
a type of noise commonly referred to as shot noise or quantum noise. This is an almost fundamental
limit1 More discussion will be given to ways to circumvent fundamental limits in sections 6.3 and 15.3. Such
circumvention, however, is not yet of practical value, although it does seem that anytime someone states a
fundamental limit, it is for their own egoto show that they are aware of fundamental thingsand has
little to actually do with science in general. on the emission/detection process, which turns out to be quite
serious for analog communications although much more benign in the digital case.
As was mentioned above, the average energy of a thermal phonon is roughly Boltzmanns constant k times
the temperature T , which for room temperature is roughly 1/40 of an eV. Optical quantum detectors can
operate at room temperature, as single photons are measurable. Therefore, optical direct detection can be
quite sensitive if shot noise-limited. Direct detection, further, is totally compatible with intensity modulation
schemesschemes in which the source is essentially just turned on and o. Such modulation schemes are the
easiest to implement. When coupled with lights short wavelength which allows for miniature sources and
detectors and micron-sized waveguides, direct detection schemes have allowed for small, lightweight, high-
bandwidth systems which are competitive in many areas, most notably to the present telecommunications
transmission, although a myriad of other applications are continually opening up. As mentioned previously,
these applications have tended to open up more slowly than originally predicted, as cost was really not much
of a consideration in telecommunications, where equipment costs are swamped by other considerations.
With consumer electronics, one need not worry about right of way or installation. At present, the cost of
connecting to personal computers (PCs) a few meters from each other optically is so expensive that ber has
not yet come to the consumer market. The high cost of the link in such a case, though, is not fundamental
but more historical. Present-day developments in millimeter core plastic is an example of a much cheaper
technology than, for example, glass ber. The costs of components to go into ber links as well as packaging
costs are presently being reduced and new applications are opening up. Some of these other applications
will be discussed as the presentation proceeds.

1.2 An Optical System Model

A schematic depiction of the organization of an archtypical optical system is given in Figure 1.1. The model
used here, as we will see in the subsequent discussion, is quite general and applies to practically all optical
systems that exist, including communications as well as sensing systems. For example, the earliest types
of optical communication systems were probably ones in which a lamp was sequentially hidden from sight
for xed periods between periods of being in plain view. Such a system was employed by Paul Revere in
1776, but it is very likely that such systems had been in operation since the times of the early Greeks or
Egyptians. In such a system, the information source is the lampholder, the encoder is his arms together with
the occluding sheet, the receiver optics are the eyes of the beholder, the detector is the beholders retinae,
the receiver is the brain of the beholder, and the information out is the response of the beholder. One could
say that such a system employs digitally encoded on-o keyed intensity modulation with direction detection
(a DEOOKIMDD or DIM system, depending on the receiver design) if one were to classify this system. A
similar block diagram description could apply to an Indian on a hill sending smoke signals back to his tribe,
a ber optic gas detection system, a telescope, or, as we will see as we go on, practically any optical system
we can devise.
Perhaps a major point in having a model such as in Figure 1.1 is that such a model is so general that
we can use it as a basis for comparing widely diering systems that at rst may appear incomparable.
Certainly, however, any two systems can be compared by comparing the data launched into the system
with that exiting the system (bit error rate (BER) counting in digital systems or noise gure measurement
in analog systems). Other gures of merit can, indeed, also allow comparison of the specic blocks of the
model. Such comparison will be a major occupation of this courseto compare performance of the various
1
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 5

Figure 1.1: A schematic depiction of the organization of an optical communications system in which the
square blocks are the optical system itself and the circular blocks denote the system input and output,
respectively.

schemes for performing the various functions of the blocks of Figure 1.1. The following four subsections will
discuss rst sources, then bers, and then detectors. The last section will give some systems architecture
perspective, as the other types of components which make up the blocks of Figure 1.1 in the archtypical
modern-day optical communication systems are very architecture-dependent. The discussion that follows
will be given from a somewhat historical perspective.

1.3 Optical Sources

A discussion of optical sources is the only subject of numerous courses. The present discussion, in comparison
with a full course, must by nature be cursory. In the following discussion, we will limit ourselves to sources
and salient features of those sources which relate to modern-day communications systems. (This topic is the
subject matter of those numerous courses as well.) A denition for a modern-day optical communications
system by nature must be quite arbitrary. A lighthouse beacon comprises the transmitter end of an optical
communication system but will not be considered here. Attention here will be placed on compact sources
which can be used to transmit at least MHz information rates. Such sources are the ones of interest for
telecommunications, data communications, and radio-frequency remoting and sensing. (Sensor systems have
bandwidth requirements similar to those of telecommunication systems for other reasons, among them that
phase dithers are necessary to place the beat note outside the 1/f laser noise peak (to be discussed in
section 8.4 of Chapter 8), which may extend to many MHz.)
Perhaps the important milestone date in dierentiating modern day from early times is the demon-
stration of the laser in 1960 (Maimon, 1960). One of the most important characteristics of the laser is its
temporal coherencethat is, the fact that the laser linewidth is a tiny fraction of its center frequency. Even
early lasers (HeNe for example), when operated in a single temporal mode, could exhibit relative bandwidths
(that is, linewidth divided by center frequency) of 1012 . In the early 1960s, no other optical source, even
with strong ltering, could exhibit relative bandwidths smaller than 104 or so. But this spectral purity was
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 6

Figure 1.2: Depiction of an archtypical 1960s-style optical transmitter.

not the important laser characteristic when it came to optical communications. Much more important were
spatial coherence (brightness) and, eventually, radiation eciency.
The basic principle behind the laser is to lock the emissions of an ensemble of excited atoms in a cavity
into a single direction, thereby locking the individual emissions in both frequency and phase (to be discussed
further in section 8.1). That the emissions are locked in direction is an important point, as this locking
allows for spatial coherence; that is, the phase across the whole wave front is constant. Such a beam can
achieve the minimal beam divergence possiblethat is, the divergence required by the diraction limit. A
non spatially coherent beam will diverge at many times this limit. Of course, spatial ltering can be used
to eliminate rays which refuse to stay in phase with the beams center of gravity. However, this ltering
will require an energy loss, as power must be extracted by this passive lter. Therefore, any directivity
gain comes at a price. As it turns out, this price is quite high. The second law of thermodynamics tells us
that we cannot decrease entropy without adding energy. As it turns out, a measure of the entropy of an
electromagnetic wave is the inverse of its brightness, which is dened as the amount of power that the source
can radiate from a xed area into a given solid angle. The second law of thermodynamics, therefore, requires
that passive spatial ltering cannot increase a sources brightness. Clearly, a spatially coherent source will
have the greatest brightness for a given power output as it radiates into the minimum angle. An isotropic
source would need to radiate roughly 104 times the power radiated by a laser with a 1 milliradian beam
divergence in order to have an equivalent brightness. Further, as the diraction angle is directly proportional
to wavelength, a diraction limited microwave source would need to radiate roughly 105 times the power of a
diraction limited optical source in order to achieve the same brightness. In communications, one generally
wants to send information from one point A to a number of points B, C, D, . . . , and not to disperse the
transmission throughout all space (broadcast). It is clear from the above discussion that a coherent optical
source will be the most ecient source of all for achieving this goal.
Figure 1.2 depicts a 1960s-style optical transmitter. During this early era of modern-day optical
communications, the available laser sources were either gas or solid-state lasers. These early lasers were also
characteristically of low gain per unit length in their cavities and therefore needed to have high Q (quality
factor) cavities. Essentially, a high Q means a highly resonant structure which is also a structure which
by nature must store energy for a period of time which is long compared to an optical period. A problem
with such structures is that of pumping; that is, it is hard to change the state of the elds within the cavity
without waiting for many roundtrip times of the cavity. The cavity has, so to speak, a built-in memory
of its past state. For this reason, output stability would require stability of the laser pump, and clearly
modulation of the output light could not be achieved by pump modulation but only by external modulation.
This external modulation would therefore require enough optics to collimate the laser beam, focus it through
a modulating crystal, and then recollimate the beam.
Two major developments occurred in 1970 that greatly altered the situation in optics in general and in
optical communications in particular. The rst was the development of the laser diode which could operate
at room temperature (Hayashi and Panish 1970, Alferov et al 1970). The original semiconductor laser diode
was developed in 1961 (Basov et al 1961, Hall et al 1962, Nathan et al 1962) but required such high current
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 7

Figure 1.3: Schematic depiction of an optical transmitter which employs direct modulation of the laser bias
current Idc by directly summing Idc in a bias T with an information current i(t).

densities to achieve lasing threshold that low-temperature operation was required. (More recently, a dierent
conguration of laser diode, the surface-emitting laser diode or vertical cavity laser diode (as opposed to the
original edge-emitting laser diode) has appeared.) However, even these early laser diodes had some striking
properties with respect to other lasers. As very high carrier densities are obtainable in semiconductor diode
junctions, the gain per length in a laser diode can be very high with respect to, for example, a gas laser
medium. For this reason, laser diodes could be made to operate in very low Q, short cavities. Now, a high
Q (or high nesse) cavity requires both long length and high reectivity cavity mirrors, whereas the laser
diode cavity could be short (circa 300 m) and have low reectivity mirrors (actually, cleaved semiconductor
facets with reectivities on the order of a few percent). For this reason, the laser could be modulated
through manipulation of the current owing through its junction, thereby obviating the need for an external
modulator. In a high Q resonator, changes in input power or current would cause a long-lasting ringing
which can completely distort any impressed information. An optical transmitter which takes advantage of
this direct current modulation characteristic is illustrated in Figure 1.3. It should be noted here, however,
that a close cousin of the laser diode developed along with it during this periodthat is, the light-emitting
diode (LED). A light-emitting diode has essentially the same structure as a laser diode but has essentially
zero reectivity facets and therefore never reaches a lasing threshold. The light emission, therefore, can
never become as spatially coherent in an LED as it is in a laser diode. However, due to the small size of
the junction and the fact that, due to the nite depth of the device some coherent emission could occur,
the emitted light is signicantly more bright than that from totally incoherent sources. Indeed, the LED
can be directly modulated as the laser diode. Further, the LEDs are extremely inexpensive to manufacture.
For these reasons, the LED showed up as a source in many transmitter modules such as the one depicted
in Figure 1.3, when system requirements were not so stringent as to require a laser source. A disadvantage,
however, of the LED is wall plug eciency. A laser diode can be close to 100% ecient. As prices drop for
laser diodes of all types, they tend to replace LEDs due to the better collimation and wall plug eciency.

1.4 Optical Paths

The second major development to occur during 1970 was the development of the low loss optical ber
(Kapron, Keck, and Maurer 1970). The idea of communicating through free space is well and good, but as
was previously stated, a major advantage of optical transmitters was their directivity. For broadcasting, it is
hard to beat radio waves. For point to point, it is hard to beat optics. But there are two problems with using
free space as a transmission medium. The rst one is very practical. Usually, one is trying to communicate
with someone else on the face of the Earth. (Satellite-to-satellite communications will be briey discussed
in some following chapters, in particular 12.1 of Chapter 12 on digital communications, where pulse position
modulation (PPM) is mentioned.) Two major problems arise from this. One problem is that the Earth is
round. Once line-of-sight distances are exceeded, one needs to increase link length by a quantum leap in
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 8

Figure 1.4: Schematic depiction of an optical transmitter as used in the telephone network.

order to include a satellite repeater. This is a problem for both microwave and optical transmission. As
was shown by Hertz (Hertz 1983), although the rst denitive trans Atlantic demonstration was made by
Marconi some years later around the turn of the century, low-frequency waves (AM band, 1 MHz) will
cling to the ground for some distance. Already by the shortwave band ( 10 MHz), the waves begin to skip
o the Earth, although up to roughly 100 MHz they still reect o the ionosphere. At higher frequencies,
one needs an orbiting reector and/or repeater. Another problem with free-space optical transmission is
that, unfortunately, there is a thing surrounding the Earth called the atmosphere. Radio waves dont care
too much what is happening in the atmosphere, but optical waves do. Rain, snow, fog, and even wind aect
optical transmission. There are free-space optical links still in use, especially between buildings in cities and
on campuses, but for any but the shortest, most protected of these links, weather interference does occur.
Many sensors, though, by nature, use free space and always will as they are measuring the weather. An
example is the LIDAR, or laser radar. (Attention will turn to communication system models of some simple
sensor systems in Chapter 14.) A third problem with free-space communication is more fundamental. That
problem is diraction. Coherent waves in free space will expand at an angle that is roughly equal to the
wavelength of the radiation divided by the eective radiating aperture. One can minimize the diraction
eect only by using larger and larger focusing lenses. In fact, one can project a 600 m spot on the moon, but
this requires using a 2.7 m telescope as the transmitter. The diraction eect, therefore, puts fundamental
bounds on distances and powers necessary in free-space systems. The optical ber, however, is a solution
to both of the above-mentioned problems, at least in commercial telecommunications systems and in some
sensor systems as well.
A modern-day archetypical telecommunications optical transmitter, such as the one employed in todays
telephone network, is depicted in Figure 1.4. The idea is that the laser diode can be pigtailed with an
optical ber, therefore obviating the need for any focusing optics whatsoever. The transmitter module
therefore needs no alignment. One need only hook up the laser to a current source and hook up the ber
output into a transmission ber by means of an optical connector.

1.5 Optical Detectors

As was discussed previously in Section 1.3, the photon energy is both a boon and a hindrance to optical
communications. That the photon energy is much larger than the room temperature phonon energy, kT ,
is advantageous, as that means that techniques can be found which are extremely quantum ecient. This,
when coupled with the fact that the photon energies correspond closely to various atomic transitions and
semiconductor bandgaps, allows for a plethora of dierent materials to serve as detector elements, each
being capable of detecting even single photons. At the same time, however, this high photon energy yields
a limitation. As photons are energetic, in order to receive a small number (low-power signal in a short
interval), one must put up with a signal granularity; that is, a major source of noise will be counting noise.
The high energy per photon means also that there is an upper limit on how many photons a detector can
receive before it is fried from residual thermal energy. This form of counting noise, often referred to as
shot noise or quantum noise, forms the technical noise oor in optical communications. Although techniques
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 9

Figure 1.5: Schematic depiction of the workings of the eye.

to circumvent this oor exist (as will be touched upon in sections 6.3 and 15.3), they are not yet viable as
practical signal control and processing techniques.
The earliest known and one of the best light detectors is the eye. The workings of the eye are illustrated
in Figure 1.5. The retina is a lens whose focal length is controlled by optic muscles. This telephoto lens
images an object onto a photosensitive plane called a fovea. Image inversion is performed in the brain. The
photosensitive plane of the eye is quite reminiscent of another, less ecacious, but presently quite popular
light detector, that of the photosensitive plate or lm. The basic workings of light-sensitive lm are given
in Figure 1.6. In the exposure phase, incident photons break bonds (often silver halide bonds). In the
development stage, the lm is immersed in a developing bath which contains chemicals which react with the
broken bonds to change the nature of one of the bond holders, generally changing a bonded molecule with
a dielectric character into one dielectric and one metallic piece. These metal clumps then act as absorption
centers, so when reilluminated they form a negative image of what was incident on the lm during the
exposure phase. Photographic lm is very useful for storing large amounts of data; however, development
tends to be time consuming, the process tends to not be photon ecient, and it is hard to get very good
resolution. It seems at present that electronic imaging arrays reading out into CD ROMs may take over as
the recording elements of the next generation of photographic instruments. We will not concern ourselves
further with imaging technology, however, in what follows.
It was Einstein in 1905 [3] who rst explained an eect known as the photoelectric eect earlier observed
by Phillip Lenard. Lenard shone light on a metal electrode in a vacuum tube and noted that the current
was proportional to light intensity if and only if the light were of certain colors. The idea is as illustrated in
Figure 1.7. If the photon energy is suciently high (exceeding the work function of the metal), an electron is
ionized from the anode, and if the electric eld between the two electrodes is suciently high, this electron
is accelerated toward the cathode, where it causes a current ow in the external circuit. Such vacuum tube
realizations of optical detectors were the most common ones until the 1960s, when solid-state semiconductor
detectors became of suciently high quality to supplant vacuum tube technology.
A basic circuit for use with a semiconductor p-i-n detector is depicted in Figure 1.8. A basic p-n diode
structure, which would develop a depletion region under reverse bias, is modied slightly to include an
intrinsic (undoped) region between the p and n regions to enlarge the depletion region suciently that it
becomes relatively insensitive to reverse bias level and therefore, for example, will not change dimension under
moderate illumination. Illumination of the intrinsic region will ionize valence electrons, thereby creating a
number of free electrons in the conduction band and holes in the valance band, where the electrons and
holes will be swept out of the junction region by the strong elds there. This sweeping out of the electrons
and holes will lead to current ow in the external circuit and will thereby temporarily lower the device bias.
These temporary lowerings are then picked up as an ac variation on the load resistor RL . The p-i-n structure
can be a high-speed structure, but, as with all high-speed structures, it must also be low-power. But it is
just to such high-speed, low-power structures that we wish to limit our attention. The PIN will be discussed
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 10

Figure 1.6: Schematic depiction of the working of a photographic lm: (a) exposure, (b) development, and
(c) the resulting lm.

Figure 1.7: Schematic depiction of a vacuum tube diode photodetector.

CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 11

Figure 1.8: Schematic depiction of a p-i-n detector and its associated bias circuit.

further in section 4.6 and will be used as an archetypical square law detector in Chapter 7 and much of the
presentation following Chapter 7.
The other detector (in addition to the p-i-n) in common use in optical communication systems is the
avalanche photodiode (APD). The APD has an operating principle similar to that of the p-i-n, but it has
a much longer propagation path for the electron. By applying a high voltage along this path, the electron
impact ionizes a multiplicity of additional carriers, leading to an eective propagation gain. A drawback
with APDs was the need for kilovolt voltage supplies, although the situation is improving. There is always
a noise penalty, as will be discussed in section 11.4.2.

1.6 Some System Considerations

The demonstration of both the low-loss optical ber and the room-temperature semiconductor laser was
timely, as the need for a new transmission medium for the telephone network was already clear. The original
telephone network appeared much as the one depicted in Figure 1.9. (See, for example, Talley 1979 or, for a
more ber optic perspective, Personick 1993.) Central oces, from which the local (subscriber) loops which
extended to customers originated, were hooked together by trunk lines of up to a few kilometers in length.
These local branch oces would all be linked to a toll oce, which might be citywide in smaller cities. These
toll oces were the gateway to the long distance network and, through a system of protocols, could hook a
subscriber up with the next town, next state, or foreign country. These long distance oces would be linked
to each other by much longer interconnects which in this early era were generally radio links. Already by
roughly 1960, network planners began to realize that the existing trunk rights of way would eventually be
lled with cable, and further expansion of the telephone network would be stopped, as with copper cable, to
the lowest order, the aggregate information rate was only a function of the total copper cross-section. During
the 1960s, however, digital electronics became cheap and began to become advanced. Switching functions
which were originally performed manually or mechanically could now be performed by electronics much more
compactly and at much higher rates. Time-division multiplexing (TDM) became a realistic technique for
minimizing the horrendous number of cables that linked local branch oces together. However, the higher
rates did not help as long as copper was the medium.
The rst pre-ber eort at achieving large-scale TDM was the one in which a standard DS1 transmission
rate, which consisted of sixteen interleaved telephone conversations (composite rate of roughly 1.54 Mbps),
was dictated to be the trunk rate. That is, local oces would still talk to the local loop on multiple single
conversation cables, but interlocal oce communications would always be multiplexed. The reasoning for
this above all else was to minimize the amount of stu in the transmission ducts, where capacity was
rapidly being exceeded. The original standard transmission scheme to perform this function was called
T12 The nomenclature T1, although originally coined to refer only to one specic means to transmit a DS1
signal, has taken on a life of its own and is now often used synonymously with DS1. and consisted of a
cable transmission line with a 2 km repeater spacing. The problem with this T1 scheme was that it was not
expandable as was hoped. That is to say, if one were to raise the rate (the number of channels to be TDMd),
2
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 12

Figure 1.9: Schematic depiction of a highly simplied telephone network showing the typical path from
a subscriber (in the local loop) through central oces (for local calls) or to a toll center to enter a new
hierarchy in the long lines system. LL is local loop, CO is central oce, and TC is toll center.

the copper cable transmission medium, due to signal dispersion, loss, and crosstalk, would require a reduced
repeater spacing. Changing equipment in the local oce is one thing, but going into the ducts is expensive.
Further, if expansion required ever more repeaters, the ducts would eventually ll with repeaters. This
situation was threatening to shut down the growth of the national telephone network as of the late 1960s.
Although the T1 scheme was found to be (only theoretically at that time, but possible today) extendable to
DS1C (=2 DS1) but not beyond, the T1 scheme was considered not suciently expandable, and the search
for alternative transmission media became frenzied. In the last few years of the 1960s before the low-loss
ber was demonstrated, a myriad of technical systems were researched including a complex millimeter wave
waveguide technology and, I understand, even waveguides made from ice.
Originally it was multimode ber that oered the solution. These original bers had 50-m cores and
125-m outer cladding diameter and had numerical apertures of 0.2. Dispersion was such that 10s to 100s
of MHz could be transmitted over the 2 km repeater spacing, and the loss at the original 0.83 m operating
wavelength of roughly 3 dB/km, even including laser and detector coupling, posed no problem at all for lasers
capable of 100 W output. In fact, at the 100 W level, the link operation could even be shot noise-limited.
Multimode ber provided a greatly expandable T1 solution which could later be expanded from DS1 to DS3
(=28 DS1 =45 Mbps) and upward in data rate. The rst eld tests of multimode trunk links were installed
in 1975 and were a great success, and by 1980 this technology was the chosen one.
The original predictions were that ber technology would rapidly extend into the local loop. This has yet
to happen, in part due to a regulatory morass and in part due to plain economics. The ber actually took
o in the other directiontoward the long haul. In many senses, single-mode ber is a simpler technology
than multimode, despite the more exacting tolerances. If one can simply hold the ber radius and numerical
aperture product to less than some constant in order to achieve single-mode operation, the ber automatically
will have a thousand times less dispersion than one with even two modes. Further, by moving the operating
wavelength further into the infrared at 1.3 m or 1.55 m, one could decrease the ber loss by a factor of
10 in dB. The enabling technologies, single-mode ber production, and single-mode 1.3 m laser production
were in place by 1980, and soon repeater spacings of 10 km at rates of 565 Mbps were realized. These original
single-mode bers had roughly 10-m cores, 125-m outer cladding diameter, and numerical apertures of
0.1. The Institute of Electrical and Electronic Engineers (IEEE) synchronous optical network (SONET)
standard was put in place around this time, a standard in which universal multiplexing standards had rates
ever increasing in factors of 4 (565 Mbps, 2.5 Gbps, to 10 Gbps, . . .). The standard has been quite useful
in allowing electronics manufacturers to compete on receiver and transmitter electronics, thereby greatly
lowering electronics prices. Rates and repeater spacings have ever since been steadily increasing. Twenty
kilometer repeater spacing and 2.5 Gbps rates are not at all uncommon on long-haul links, and there are
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 13

Figure 1.10: Basic interconnection topologies applicable to, for example, trunk lines between local oces:
(a) a complete interconnection; (b) a star; (c) a ring.

new transoceanic ber cables as well. There are some 10-Gbps links already in place, and work on 40-Gbps
links is progressing. A problem with these rates of greater than 2.5 Gbps in the present-day market has
become cost. With the breakup of telecommunications monopolies essentially worldwide, the companies no
longer have the capital necessary to support internal electronics foundries. As the number of components
necessary in the telecommunications network is small relative to the commercial electronics market, it is hard
to convince electronic components manufacturers to set up production lines and compete on components
which have no market yet outside the small telecommunications market.
Our communications needs, however, extend beyond Plain Old Telephone Service (POTS). These days,
cable TV as well as two-way (MODEM) computer interconnections are becoming more and more common.
Digital television (DTV) transmission, if uncompressed, requires roughly 140 Mbps for a single channel.
High-denition television (HDTV), if uncompressed, requires four times the DTV rate or roughly 560 Mbps.
Many such channels could be carried by telephone long lines but barely, if at all, by trunks and in no way
by the local loop, which has so far been little impacted by the ber communications revolution. Cable
companies provide TV to the home through use of the standard 5 MHz analog transmission. More and
more, the cable companies are achieving reliability and delity through the use of ber trunks and only
short twisted-pair feeders. MODEM computer connections are notoriously slow. Businesses which have a
great enough need can buy bandwidth through setting up their own private branch oce (PBX) and hook
into the long lines system and provide their own local distribution. Broadband services such as the graphic
interfaces to the internet, however, could nd a much greater usage could they be brought closer or even
directly into the home or small business. Although the technology is there to bring multimedia services to
small businesses and homes, the costs and regulations are hard to surmount.
A short discussion of interconnection topologies seems appropriate at present. The basic fully intercon-
nected star and ring congurations have already been matter of factly illustrated in Figure 1.9, but they are
individually broken out for study in Figure 1.10. For a conguration where there are N nodes, there are
various manners in which they can be interconnected. A maximal interconnect would be one in which each
node was interconnected with each other node, requiring each node to have (N 1) interconnects requiring
N (N 1)/2 interconnection wires, which quadratically becomes a large number of interconnection wires as
the number of nodes increases. A star interconnect, as illustrated in Figure 1.10(b), requires only N 1
interconnects total but requires a smart head end through which all messages must pass for routing and/or
broadcast. This function is distributed in the maximal interconnect where each node could serve as a server
or router. In the ring (or its duplex version known as a bus), as illustrated in Figure 1.10(c), there are again
only (N 1) interconnects, but here the processing can be either localized (in the head end) or distributed
through the ring. However, in the ring, everybodys data must pass through each node, whereas in the other
two congurations only the data necessary to reach the receiver need pass through any interconnection,
ostensibly requiring much less bandwidth.
Any communications system of any great extent is probably going to require a combination of topologies,
as was indicated in the discussions surrounding Figure 1.9, therefore becoming some kind of a tree with
round leaves perhaps located in a forest of some connectivity. The structure will probably also have built-
in redundancy to allow reprogramming to obtain diering virtual rings and busses depending on trac or
CHAPTER 1. INTRODUCTION TO OPTICAL COMMUNICATION SYSTEMS 14

down equipment. According to another IEEE set of standards, no matter what equipment is attached to
a node, each node is comprised of seven standard layers between the physical layer (network) and upper
layer (logical attachment to equipment). This standard has again allowed for reasonably complex, reliable,
cheap interconnection chip sets, allowing for a considerable degree of smartness to be built into a given
node. Even though the physical layer satises an IEEE SONET (synchronous transmission) standard, the
logical intervening layers can support asynchronous transmission; that is, the logical network due to trac
constraints or other reasons could decide to route dierent packets via dierent routes to dierent locations.
The completely interconnected bar network of Figure 1.10(a) as well as the star network of 1.10(b) can be
what are called circuit switched networks. That is, both logically and physically, either network can provide
a hard interconnect exists between sender and receiver. The bus network of Figure 1.10(c) must almost
necessarily send packets in order to identify who should receive the message. It is not truly packet-switched,
though, as there is only one way around the bus, and generally one tries to design a bus so as to remove
packets after a single traversal of the bus. A network such as that of Figure 1.10(a), however, could be packet
switched. That is, if a node were busy, the packet could go elsewhere and come back. That is, if a receiver
were busy, he could send a packet back into the network to receive it later. In fact, by logical operations
alone, the network of Figure 1.10(a) could recongure itself into either Figure 1.10(b) or Figure 1.10(c),
depending on what it wanted to do. The packet switching concept, though, becomes more powerful with
system complexity. It is packet switching which makes the World Wide Web possible.
A major point of the above discussion is to point out where ber optics per se may serve to really change
telecommunications. To a great degree, at present, bers have been used to replace wires or radio links
one-for-one without signicantly aecting network function. Now a change is in the wind. As was pointed
out earlier in this introductory chapter, optics is a good technology for space, time, and wavelength division
multiplexing (SDM, TDM, and WDM), making it maximally exible. In the past, a major reason for using
stars for both local area networks (LANs) as well as distribution from the local loop was that the star
minimized the number of interconnects of low bandwidth. With ber, one needs to minimize neither the
bandwidth nor the number of interconnects. Further, with smart nodes, one can use sophisticated routing
schemes as long as the node has enough bandwidth to match that carried by the ber. For these reasons,
the coming generations of trunk networks will be very high-speed (2.5 Gbps to 10 Gbps to 40 Gbps) local
loops which should allow broadband services to come a step closer to the home, if not yet to the local loop
and home. A review article by Personick (Personick 1993) points to some directions the telephone system
will take, including some discussion of bringing broadband to the trunks. It should be noted here that sensor
systems can also prot from combinations of SDM, TDM, and WDM, and this is a prime reason for the
steady growth in optical sensing in general and ber and integrated optic sensing in particular. We will give
some further discussion to sensors in Chapter 14.
In the chapters of Part III of this text, the basic principles of optical communications systems will be
elucidated, hopefully in such a manner that the reader upon completion of the material will be able to analyze
and design future optical systems. At the least, he or she should be able to study the various applications
discussed in Part IV of the text and be able to return to the introduction and read through it with better
understanding than during the rst run-through.
Bibliography

[1] Zh. I. Alferov, M. V. Andreev, E. L. Portnoi, and M. K. Trukan, AlAs-GaAs Heterojunction

Injection Lasers with a Low Room-Temperature Threshold, Sov Phys Semiconductors 3, pp. 1107
1110 (1970).

[2] N. G. Basov, O. N. Krokhin, and Yu. M. Popov, Production of Negative Temperature States in
P-N Junctions of Degenerate Semiconductors, J E T P 40, pg. 1320 (1961).

[3] A. Einstein, Uber einen die Erzeugung und Verwandlung des Lichtes Betreenden Heuristischen
Gesightspunkt, Annalen der Physik 17, series 4, pp. 132148 (1905).

[4] R. N. Hall, G. E. Fenner, J. O. Kingsley, T. J. Soltys, and R. O. Carlson, Coherent Light Emission
from GaAs Junctions, Phys Rev Lett 9, pg. 366 (1962).

[5] I. Hayashi and M. B. Panish, GaAs-Gax Al1x As Heterostructure Injection Lasers Which Exhibit
Low Thresholds at Room Temperature, J Appl Phys 41, pp. 150163 (1970).

[6] H. Hertz, Electric Waves (MacMillan, 1983). This is a reprint of an 1887 original of Hertz.

[7] F. P. Kapron, D. B. Keck, and R. D. Maurer, Radiation Losses in Glass Optical Waveguides,
Appl Phys Lett 17, pp. 423425 (1970).

[8] T. H. Maimon, Optical and Microwave-Optical Experiments in Ruby, Phys Rev Lett 4, pp. 564
565 (1960).

[9] M. I. Nathan, W. P. Dunke, G. Burns, F. H. Dills, and G. Lasher, Stimulated Emission of Radiation
from GaAs p-n Junctions, Appl Phys Lett 1, pg. 62 (1962).

[10] Stewart D. Personick, Towards Global Information Networking, Proc IEEE, 81, pp. 15491557
(November 1993).

[11] David Talley, Basic Telephone Switching Systems, Second Edition (Hasbroughk Heights, NJ: Hay-
den, 1979).

15