dvances in solid-state and photonic tech-nologies have made the “can’t be done”

concept a thing of the past. Bit rates of2.5 Gb/s, 10 Gb/s, and 40 Gb/s over

many kilometers of single-mode fiber are a real-ity. The driver for this bandwidth appetite wastriggered with early optical networks (SONET/SDH) that demonstrated that glass fiber is atransmission medium that permits light to travelthrough it without amplification for hundreds ofkilometers and at incredible data rates (manyGb/s), two accounts that were not possible withcopper-twisted-pair cable. And this appetite hasbeen accelerated ever since with the proliferationof transferring additional information over thecommunications network with major servicecontributors such as e-mail, e-commerce, theInternet, electronic documentation transfer,video, and ubiquitous mobile telephony. And thisis just the beginning as more “exotic” servicesare planned for and contemplated to be offeredover the communications network. This articlediscusses dense wavelength division multi-plexing (DWDM) photonic technology and therole it will play in shaping future communica-tions networks.

What Is DWDM andHow Does It Work?Bandwidth Elasticity

As bandwidth demand keeps increasing, how canwe be assured that the network is elastic enough

� 8 8755-3996/02/$17.00 ©2002 IEEE IEEE CIRCUITS & DEVICES MAGAZINE � JANUARY 2002

Stamatis V. Kartalopoulos

©E

YE

WIR

E;1

999

AR

TV

ILLE

LLC

,200

1A

RT

VIL

LELL

C;2

001

IMA

GE

-ST

AT

E

to cope with this in-crease? How do we as-sure that systems andnetworks are able to pro-cess and transport an in-creasing volume of voiceand data (video,high-speed data, inter-active multimedia, etc.)traffic? How do we as-sure that the network isable to manage a contin-uously increasing bandwidth (as new flexible services becomeavailable, such as bandwidth on demand)?

Currently, there are technological choices to answer thesequestions:� Install more and better fiber, as the need arises. Although

this is currently pursued, nevertheless, it requires substan-tial planning and investment, and it may not be possible inall cases.

� Use higher speed photonic technology to increase the bitrate (to 10 Gb/s, 40 Gb/s, > 100 Gb/s). This implies that tostay at the forefront of data rate, one has to always use themost advanced technology that is neither mature or costattractive.

� Use optical components instead of electronic components(e.g., amplifiers, filters, etc.). This is a design choice thatdepends on the availability of a range of components andhow well their specifications match. However, becausemost optic components are data-rate independent, the costper bit turns out to be a tiny fraction compared with elec-tronic implementation.

� Increase the number of optical carriers (wavelengths) persingle fiber, a technology known as wavelength divisionmultiplexing (WDM). This is a successful technology as ittakes advantage of existing fiber, and the only changes re-quired are at its termination points; in some cases, itgreatly simplifies the traditional regeneration as opticalamplification is much simpler and more cost effective.

In all these choices, photonic technology plays a pivotal rolein the communications network. In particular, network band-width elasticity is better addressed with WDM.

Depending on network application, WDM comes in two fun-damentally different flavors, each with its own complexity, speci-fications, and cost structure: dense-WDM (DWDM) with morethan 80 wavelength channels per fiber, and coarse-WDM(CWDM) with less than 40 wavelength channels per fiber. How-ever, as time progresses and technology permits, we will witnessa shift of the demarcation points, and DWDM will have morethan 200 (and perhaps 1000) wavelengths, whereas CWDM willhave more than 40.

Wavelength Division MultiplexingEarly optical transmission in long-haul single-mode fiber appli-cations used a single wavelength at 1310 nm, whereas wave-

lengths in the 800-nmband have been used inshort-haul multimodefiber applications. How-ever, photonic devicesperforming in thelow-loss spectral band ofabout 1.55 µm have en-abled more than a singlewavelength in the samefiber (Fig. 1). Having puta number of wave-

lengths in the same fiber, the aggregate bandwidth per fiber ismultiplied by this number. For example, 40 wavelengths at 40Gb/s each yield an aggregate bandwidth of 1.6 trillion bits persecond per fiber (1.6 Tb/s). In SONET terms, this is equivalent to20 million simultaneous conversations per fiber.

Thus, DWDM technology exhibits an inherent flexibility:each wavelength is a huge transporting vehicle and it does notrecognize the type of information it carries, nor does it care [1].For example, one wavelength may carry Internet [2] and anotherSONET, or ATM [3-5]. The sender and the recipient alone handlethe content cargo appropriately. Moreover, WDM technologywith optical devices that provide functionality such as add-dropmultiplexing can transform any type of network (mesh, ring, orstar) into a physical single-fiber ring network (Fig. 2).

Based on the low-loss behavior of glass fiber (Fig. 3), ITU-Thas defined a grid of 81 frequencies in the C-band (196.10-192.10nm) starting with a frequency at 196.10 THz (λ = 1528.77 nm),the remaining standardized frequencies are computed decre-menting/incrementing by 50 GHz (0.39 nm). For 40 frequencies

IEEE CIRCUITS & DEVICES MAGAZINE � JANUARY 2002 9 �

1

2

N

1

2

N

λ1

λ2λN

OptAmpl.

Fiber

Diff

eren

t Ser

vice

s Different S

ervices

Transmitters

ElectronicRegime

PhotonicRegime

ElectronicRegime

Mux

Demux

Receivers

1. Fiber can support a large number of wavelengths, each carryingdifferent payload in the same fiber.

Node B

Node BNode CNode C

Node E

Node D

Node A/Hub

Node A

Node D

RingRing

Hub

Fully Connected Star Topology

2. A fully connected topology and a star topology converted into aphysical DWDM ring topology.

DWDM technology exhibits an inherentflexibility: each wavelength is a hugetransporting vehicle and it does notrecognize the type of information it

carries, nor does it care.

on the grid, the startingfrequency is the samebut the decrement/in-crement is 100 GHz, andfor 20 frequencies it is200 GHz. Decrementingby 50 GHz beyond theC-band results in 80more wavelengths de-fined in the L-band. Sim-ilarly, when incrementing by 50 GHz, more wavelengths aredefined in a band that requires specialized single-mode fiberwith low loss characteristics in the 1.4-nm band where HO- has

been causing high ab-sorption. Table 1 liststhe wavelength bandsused in optical commu-nications.

The scalability ofDWDM does not stophere; research has dem-onstrated that the num-ber of wavelengths per

fiber could increase to more than 1000, and this clearly is not alimit. Bandwidth at this level would enable every single person inthe United States to be continuously connected with anyoneelse—not only with voice but also with video and data. Moreover,considering that a cable could have more than 200 fibers lighted,one realizes that DWDM enables not only every single person inthe United States to be continuously connected but also, per-haps, every single person in the world.

DWDM Technology EnablersPure DWDM systems are supposed to be “all-optical.” That is,functionality that was previously implemented with electronicsis now achieved with all-optical devices, such as those listed inthe following sections.

FiltersFiltering is accomplished with passive Fabry-Perot, Bragg,thin-film, Mach-Zehnder, dielectric, and acousto-optic filters.Each filter is based on different principles of physics; for exam-ple, the Fabry-Perot is based on interferometry, the Bragg on dif-fraction, and the prisms on refraction.

Among the most significant are the Bragg gratings as theyare passive devices, easily manufactured, easily integrated withother components, and cost effective. Diffraction gratings comein different flavors: reflected, pass-through, fiber, and so on.However, the diffraction theory is the same, and it may differ inonly some parameters differentiating the plate from the fibergratings. In all, the condition for strong reflection, also known asthe Bragg condition, is:

d = −mλB/2

� 10 IEEE CIRCUITS & DEVICES MAGAZINE � JANUARY 2002

Sili

ca F

iber

Atte

nuat

ion

(dB

/km

) 0.6

0.5

0.4

0.3

0.2

0.11100 1200 1300 1400 1500 1600 1700

Wavelength (nm)

Attenuation Due to OH–

EDFABand

3. Loss performance of glass fiber.

Table 1. Frequency Bands in Optical Communications

Window Label Range Fiber Type Applications

1st – 820-900 nm MMF

2nd S 1280-1350 nm SMF Single-λ

3rd C 1528-1561 nm SMF DWDM

4th L 1561-1620 nm DSF DWDM

5th – 1350-1450 nmSMF

AllWave™DWDM

Pure DWDM systems are supposed to be“all-optical”—functionality that was previouslyimplemented with electronics is now achieved

with all-optical devices.

λ1 λ1λ3 λ3λ4 λ4λ2

λ2d

UV Interference Pattern

Fiber

GratingCladding Core (n)

4. When the fiber core is exposed to a UV periodic pattern, the refractiveindex is affected permanently and a fiber Bragg grating is constructed.

DiffractionGrating

Fibers λ1

λ2

λN

Θ

Lens

Incident Beam + + ...λ λ λ1 2 N

DiffractedWavelengths

5. A grating-based optical multiplexer.

where m is an integer, dis the grating constant,and λB is the wave-length for which theBragg resonates and re-flects.

Figure 4 illustrates afiber Bragg grating. Ad-vanced fiber Bragg grat-ings, known as chirpedBragg gratings, havealso the additional abil-ity to compensate for chromatic dispersion.

Multiplexers and DemultiplexersWavelength multiplexing and demultiplexing is accomplished withpassive components such as diffraction gratings, thin-films, andsuper-prisms. For reflected and pass-through gratings, the angularseparation between wavelengths for a given order m, also known asthe angular dispersion D, is:

dβ/dλ = m/(dcosβ)

where m is an integer, d is the grating constant, and β is the an-gle of diffraction.

Figure 5 illustrates the principles of a diffraction-baseddemultiplexer.

Optical SwitchingSwitching is accomplished with solid-state technology (lith-ium-niobate), micro-electro-mechanical mirror systems(MEMS), tiny bubbles, liquid crystals, electro-holographicmethods, and other technologies that are still in the experimen-tal phase. However, each technology has its own merits and eachone finds its own niche in the applications space. For example,MEMS make large optical cross-connecting fabrics (1000 ×1000) but are slow switching (ms) compared to lithium niobate,

which makes small fab-rics (32 × 32) but is faster(ns). Table 2 lists a com-parative sample of opti-cal switching tech-nologies.

Optical Add-DropMultiplexing

In communications,dropping and adding oneor more channels is an

important function that permits efficient bandwidth delivery anddistribution over the network. Optical add-drop multiplexing isaccomplished by combining optical demultiplexers andmultiplexers, optical switches, filters, and other components.Figure 6 illustrates a single wavelength optical add-drop multi-plexer that employs a Bragg fiber grating.

Employing tunable components, dynamic add-drop wave-length multiplexing enables dynamic wavelength assignment,dynamic bandwidth allocation, service and network protectionand survivability, and overall great flexibility in cost-efficientbandwidth management.

Optical AmplificationDirect optical amplification [6] is accomplished with specializeddoped fibers (e.g., erbium-doped fiber amplifiers (EDFA)), semi-conductor optical amplifiers (SOA), and Raman amplification.Each amplifier has its own characteristics and is usable in a dif-ferent spectrum (Fig. 7). In short-haul applications, the deploy-ment of amplification is limited, but in long-haul, where it ismost required, it is not unreasonable to use more than one am-plification method, such as EDFA and Raman combined.

Optical RegeneratorsIn long-haul optical transmission, the signal requires periodicregeneration (or repeaters) in order to reach a destination that

IEEE CIRCUITS & DEVICES MAGAZINE � JANUARY 2002 11 �

DWDM will quickly grow into a technologyfor a global communications network thatwill allow anyone at anytime and at any

place to communicate with voice, with fastdata, and with picture.

OpticalAmplifier

OpticalAmplifierCirculator

BraggGrating

Coupler-Combiner

Equalizer

I Isolator

λk λk

6. A grating-based single wavelength optical add-drop multiplexer.

may be many hundreds of kilometers afar. Traditional regenera-tion requires conversion of the optical signal into electric retim-ing, reshaping, and regeneration (amplification), a functionknown as 3R. Then, the signal is converted to optical and trans-mitted to the next repeater, and so on. However, traditional re-peaters are complex devices; they require maintenance and arecostly. Optical technology allows for direct optical amplificationthat can stretch the distance between amplifiers to many kilo-meters, and thus lower the overall cost and maintenance; this isknown as 1R, since all that the amplifiers accomplish is regener-ation. However, there are optical components and techniquesthat can also accomplish reshaping, and more recently retiming,in the optical regime (the latter is experimental). For example,dispersion compensating fiber removes the pulse widening dueto nonuniform propagation of wavelength in the glass medium

(Fig. 8). Dynamic equalization restores the amplitude of eachchannel in the DWDM signal to within a small fraction of vari-ability (~0.1%). Optical phase-lock-loop techniques retime thereceived optical pulses to remove drift and jitter. Thus, althoughthe current state of the art is 1R or 2R repeaters, all-optical 3Rsare on the research and development bench.

Other Optical ComponentsThe list of optical components and technology does not stophere. Lasers and laser pumps, detectors, wavelength converters,couplers and splitters, polarizers, isolators, equalizers, disper-sion compensators, specialty fibers, fiber couplers, pigtails, mi-cro-lens systems, pulse retimers, pulse reshapers, detectors,fixed or tunable devices, memories, and more will eventuallytransform the communications network landscape to an all-op-tical DWDM network (Fig. 9). In addition, new artificial materi-als with new photonic properties will make new additions to anoptical designer’s tool-box.

What About Electronics?At this point, we are compelled to answer an important question.As the communications network is transformed to an all-opticalnetwork, what will the role of electronics be?

In an all-optical network, the signal path between transmit-ter (laser) and receiver (photodetector) will be “optical” (domi-nated by optical components), known as the optical regime,whereas the remainder of the path (before the laser and after thephotodetector) will be fully electrical, known as the electrical re-gime, and in it electronics will be the only game. However, even

in the optical regime, there areoptoelectronic sensors tomonitor the optical path forperformance and all-elec-tronic devices to process per-formance data as well as tocontrol the behavior of (opti-cal) devices and to communi-cate with the various units of asystem and of a network. Thus,electronics will also be indis-pensable players in this“all-optical” network.

Network “Nodes”Although we use the term“node” [7, 8] in a habitualmanner (emanating from tra-ditional networks), data net-works employ “routers.”However, in DWDM applica-tions an advanced router per-forms DSn and OC-n groom-ing, optical multiplexing, andswitching, and it also providesquality of service (QoS); that

� 12 IEEE CIRCUITS & DEVICES MAGAZINE � JANUARY 2002

Sat

urat

ed O

utpu

t Pow

er (

dBm

)

40

30

20

10

0 1200 1400 1600 1800

YD

FA Raman

ED

FA

PD

FA

TD

FA

Wavelength (nm)

7. Optical amplifiers cover a wide spectral range.

SSMF DCFt t t

λ λ λ

Signal DispersedSignal

Compensated for

Longer Wavelengths Travel FasterLonger Wavelengths Travel Slower

8. Dispersion is compensated for with dispersion-compensating fiber placed in regular intervalsof standard single-mode fiber.

Channel 1 Channel 1

Channel 2 Channel 2

Channel N Channel N

LD

LD

LD

Opt.Mux

Opt.De-mux

Fiber Fiber Fiber

λ λ λ1 2 N, , ... λ λ λ1 2 N, , ...OA OA

OADM

= Laser Diode= Photodetector = Optical Amplifier

λk λk

λ1 λ1

λ2 λ2

λN λN

PD

PD

PD

9. An all-optical point-to-point network with an add-drop multiplexer.

is, all attributes and functions of a traditional communicationsnode. Similarly, traditional nodes have data (and packet) store-and-forward routing capability. In some configurations, tradi-tional nodes and routers work side-by-side to provide traditionalsynchronous and asynchronous service, voice, and data. There-fore, although nodes and routers may be conceptually different,as the network evolves, there is a convergence of functionality,and therefore, here we do not discriminate between the two.

Future DirectionDWDM continues to evolve, and work continues feverishly onmany fronts: technology, standards, and network architectures.

On the technology side, current systems support fixed wave-length assignment for each node, or they are manually reconfig-urable. However, a wide range of tunable devices is emerging [9]that will enable dynamic wavelength assignment, optimizedbandwidth allocation, wavelength and path protection, andbetter network survivability strategies. Advanced optical tech-niques will enable fault monitoring in the optical regime. Poly-mers and photonic crystals will provide improved andcost-effective photonic performance. Finally, nanotechnology,new materials, and integration of optical functionality will cre-ate miniaturized components with complex functionality andlower power.

In the standards arena, wavelength operation, administra-tion, management, and provisioning (OAM&P); DWDM faultmanagement [10]; DWDM network management [11]; latency;and quality of service are just a few examples of current intenseactivity. Optical DWDM networks are defined and deployed thatare characterized by unprecedented bandwidth capacity, band-width elasticity, reconfigurability, reliability, and survivabilityof service of the DWDM system and of the network.

DWDM is here and in its infancy, but it will quickly grow intoa technology for a global communications network that will al-

low anyone at anytime and at any place, with the same “identifi-cation number,” to communicate with voice, with fast data, andwith picture. The beneficiaries of DWDM photonics technologywill not be only the communications field but also fields such asmedicine, commerce, home appliances, and others that willshape future lifestyles and perhaps the world.

Stamatis V. Kartalopoulos, Ph.D., is president and chief techni-cal officer of PhotonExperts in Annandale, New Jersey (e-mail:kartalopoulos@sprintmail.com).

References1. S.V. Kartalopoulos, Introduction to DWDM Technology: Data in a Rain-

bow. Piscataway, NJ: IEEE Press, 2000.

2. B. Furcht, Handbook of Internet and Multimedia: Systems and Applica-tions. Piscataway, NJ: IEEE Press, 1999.

3. S.V. Kartalopoulos, Understanding SONET/SDH and ATM: Communica-tions Networks for the Next Millennium. Piscataway, NJ: IEEE Press, 1999.

4. Synchronous Optical Network (SONET) Transport Systems: Common Ge-neric Criteria, Telcordia GR-1377, Issue 2, Dec. 1995.

5. SONET OC-192 Transport Systems Generic Criteria, Telcordia GR-253, Is-sue 3, Aug. 1996.

6. Optical Interfaces for Multi-Channel Systems with Optical Amplifiers,ITU-T Draft Rec. G.692, Oct. 1998.

7. Network Node Interface for the Optical Transport Network (OTN), ITU-TDraft Rec. G.709, Oct. 1998.

8. Characteristics of Optical Transport Networks (OTN) Equipment Func-tional Blocks, ITU-T Draft Rec. G.798, Oct. 1998.

9. S.V. Kartalopoulos, “Emerging technologies at the dawn of the millen-nium,” IEEE Commun. Mag., vol. 39, pp. 22-26, Nov. 2001.

10. S.V. Kartalopoulos, Fault Detectability in DWDM: Toward Higher SignalQuality and System Reliability. Piscataway, NJ: IEEE Press, 2001.

11. Management Aspects of the Optical Transport Network Elements, ITU-TDraft Rec. G.874, Oct. 1998. CD�

IEEE CIRCUITS & DEVICES MAGAZINE � JANUARY 2002 13 �

Table 2. Switching Technologies (Specs Are Approximate)

Switch TypeSwitching Speed

(Approx.)Insertion Loss PDL Crosstalk λ-Flatness Typical Size

Fiber-Bragg

Grating~100 µs ~2 dB 0.5 dB/cm −40 db NA Up to 32 × 32

Acoustic-Optic ~5 µs ~8 dB ~8 dB −25 db ±10 dB Up to 1 × 1024

MEMS ~10 ms 3-7 dB ~0.5 dB −50 db ~1 dB Up to 1000 × 1000

Electrorefractive

Holograms~ ns ~4 dB ~0.1 dB −40 db ~0 dB

Up to 16 × 16 and

perhaps 64 × 64

LC ~5 ms 1 dB ~0.1 dB −40 db ~2 dB Up to 16 × 16

Bubble-Jet ~10 ms ~0.2 dB ~0.2 dB −50 db NA Up to 32 × 32