Basics of DWDM for Electronic Engineers

	ABOUT THE AUTHOR Dr. Nicolae Miron was appointed president of Optune Technologies in 2000, a subsidiary of StockerYale, which is responsible for the development of a new class of tunable optical filters (TOFs) for use in the all-optical networks of the future. Prior to joining StockerYale, Dr. Miron was a senior scientist at the Institute of Atomic Physics, Laser Department in Bucharest, Romania and most recently was a senior engineer with Unican Security Systems in Canada. Dr. Miron has twenty years of experience building high-accuracy laser instrumentation, electronics design, optical design, and systems design for the development of original laser measuring equipment. Dr. Miron earned his Ph.D. in laser instrumentation for high-accuracy measurements using interferometers from the Central Institute of Physics, Bucharest, Romania in 1983.

Optical communications is one of the most dynamic sectors in today's communications arena, due to optical's huge bandwidth potential. The evolution of the Internet, one of the major bandwidth-demanding applications, requires transfer capacities that will very soon exhaust non-optical approaches to communications. An optical carrier supports modulation frequencies up to 193 THz, much beyond the limits of existing non-optical technologies. A communications channel, or channel, frequently used throughout this presentation, is associated with a modulated optical carrier. Optical wavelengths, sometimes referred to as optical frequencies, are carried primarily through optical fibers that can accept simultaneously a monomode propagation of many wavelengths, with minimal interaction between adjacent signals.

The technology of combining or multiplexing many wavelengths into the same monomodal optical fiber on the transmitter side and extracting the necessary wavelengths at the receiver side is known as wavelength division and multiplexing or WDM. WDM is referred to either as coarse WDM (CWDM), where wavelength separation is in the order of 10 nm to 20 nm, or dense WDM (DWDM), where wavelength separation is less than 0.8 nm. A major challenge is to achieve a smaller wavelength or channel separation along with higher modulation rates of the optical carrier in order to increase transmission capacity. The total number of channels available simultaneously on the same optical fiber and the maximum bit-rate or modulation frequency carried by every channel determines the transport capacity of the optical fiber, which depends on available technology. Optical fibers are just the physical support carrying signals through an optical network. The network also contains some other optical and electronic elements, such as lasers to generate the carriers, modulators, optical multiplexers to combine wavelengths, optical de-multiplexers to extract wavelengths, optical filters, optical switches, optical amplifiers, optical receivers, and electronic controllers.

An optical fiber is a core waveguide, usually with a circular cross section, filled with glass. The core is surrounded by a cladding with a refractive index lower than the refractive index of the core and protected by a coating without optical properties. Light propagates inside the core, the waveguide being created by the difference between the refractive indices of the core and of the cladding. A monomode optical fiber designed for a wavelength range between 1310 nm and 1600 nm has a core diameter of about 9 µm. Ideally, the core should be perfectly circular and filled with a homogeneous glass, but in the real life the cross-section is always elliptic and the glass is not homogeneous. These non-ideal attributes make the fiber sensitive to polarization.

Non-linear effects in optical fibers occur as a result of the dependence of the refractive index n on power density P/A, where P is the total power inside the core and A is the effective area of TEM₀₀ mode in the core. The non-linear index coefficient n₂ depends on the refractive index value and on the refractive index profile of the core.

n=n₀+n₂P/A     (1)

Optical non-linear behavior of the refractive index generates some important effects due to elastic and inelastic interactions between the lightwave and the fiber core. In elastic interactions, there is no energy transfer between the lightwave and the medium of the fiber core, the entire energy of the wave remaining within the wave. The inelastic interaction appears when the energy of the wave is above a certain threshold, which produces some energy transfer between the wave and the medium of the core. In this case, the amount of the energy transferred to the core is proportional to the energy of the wave. Four-wave mixing, self-phase modulation, and cross-phase modulation are considered important for optical communications. Stimulated Raman scattering and stimulated Brillouin scattering are inelastic interactions that are also important for optical communications.

Four-Wave Mixing (FWM) appears at the interaction of one or more wavelengths, producing new wavelengths with different intensities. The total intensity of all incident and resulting wavelengths is conserved during this parametric interaction. Four-wave mixing is the optical analog of non-linear mixing of electrical signals using the non-linear characteristic of a forward-biased diode. In the optical case, the mixing is produced by the non-linearity dependence of the refractive index on the intensity of the wave. The wavelengths resulting from four-wave mixing may interfere with the modulated carriers and can degrade signals.

Self-Phase Modulation (SPM) introduces a spectral broadening of a light pulse due to the time-dependence of the refractive index during the optical pulse, which produces a temporally varying phase of the optical pulse. For a constant frequency f of the generated light pulse, the wavelength λ(P) depends on the current pulse power P(t) and the wavelength λ(P(t)) is time-dependent:

λ(P(t))=c/[f·(n₀+n₂(χ)·P(t)/A)]      (2)

where:

c is the speed of light in a vacuum,
n₀ is the linear component of the refractive index,
n₂(χ) is the non-linear index coefficient,
χ⁽³⁾ is the third-order susceptibility tensor,
A is the effective area of the mode propagating in the optical fiber,
f is the frequency of the optical pulse, and
P(t) is the total optical power in the optical fiber.

Cross-Phase Modulation (XPM) appears when two or more waves propagate inside the fiber and interact due to the non-linearity of the refractive index produced by the total power inside the fiber. This interaction comprises many phenomena, such as FWM, harmonic generation, SPM, Raman scattering, and Brillouin scattering.

Stimulated Raman Scattering (SRS) appears for very intense pump waves above the Raman threshold RT, which is in the range of 1W. SRS generates the Stokes radiation with wavelength λs that propagates dominantly in the forward direction with respect to the pump and carries most of the energy of the pump wave. The energy difference between the pump and SRS waves propagates into the fiber as an acoustic wave. SRS can have a useful effect by turning the optical fibers into Raman amplifiers. SRS can have also a harmful effect by transferring energy from lower channels to higher channels, contributing to the crosstalk between channels. You can minimize the crosstalk by carefully selecting the pump wavelength λp to match the Raman-gain spectrum of the fiber within the wavelength range of the channels to be amplified. Another detrimental effect of using high pump power to produce Raman amplification is to bring the fiber core into optical non-linear operation, creating conditions that result in elastic interactions such as four-wave mixing, self-phase modulation, and cross-phase modulation.

Stimulated Brillouin Scattering (SBS) appears above the Brillouin threshold BT, about 1 mW for a pulse width of the pump greater than 1 ms. If the pulse width is shorter than 10ns, SBS nearly ceases to occur. The Brillouin radiation propagates dominantly in the backward direction with respect to the pump, having a frequency shift in 10 GHz range, about three orders of magnitude smaller than the SRS frequency shift. The energy difference between the pump wave and SBS wave propagates into the fiber as an acoustical wave.

An optical fiber as a waveguide has propagation properties that depend on wavelength due to the dispersive properties of the core and of the glass filling the core. Both these wavelength dependencies are expressed by β(f), the mode propagation function into fiber. There are some parameters frequently used in optical communications:

• The first derivative of β₁(f) is the group velocity v_g.
• The second derivative of β₂(f), the group velocity dispersion (GVD) parameter.
• Chromatic dispersion coefficient (dispersion parameter) D, expressed in [ps/(nm·Km)].
• Dispersion slope S, expressed in [ps/(nm²·Km)].

There are three types of single mode optical fibers with different chromatic dispersion properties specified by the International Telecommunication Union (ITU):

• Dispersion-unshifted fiber specified by the ITU-G.652 recommendation, with a dispersion coefficient D of about 3.5 ps/(nm·Km) for a wavelength interval centered at 1300 nm.
• Dispersion-shifted fiber specified by the G.653 recommendation, with a minimum value of D of around 3.5 ps/(nm·Km) for a wavelength interval centered at 1550 nm.
• Non-zero dispersion-shifted fiber specified by the G.655 recommendation, with D between 0.1 ps/(nm·Km) and 6.0 ps/(nm·Km) for a wavelength interval between 1530 nm and 1560 nm. This attribute makes the fiber less sensitive to polarization-mode dispersion and to four-wave mixing in long-haul communication.

Issues related to polarization in monomode optical fibers occur for two reasons:

• Linear polarization of the light beam generated by semiconductor lasers
• Birefringence of the fiber coming from the optical anisotropy of the fiber core related to the departure of the core from a circular shape and to the local anisotropic properties of the core's refractive index.

The birefringence of the fiber broadens an optical pulse and changes its shape during propagation, a phenomenon called polarization mode dispersion (PMD), which, added to chromatic dispersion, limits transmission speed in long-haul communications.

Transmitters: Lasers
Only single-mode semiconductor lasers generate a well-defined wavelength that you can use in optical communications as an optical carrier. These lasers have some features similar to the features encountered at carrier generators used in radio communications, such as stable single-frequency output, modulation capabilities in the gigabit/second range, and stable operating lifetimes in the range of 1 to 100 million hours at room temperature. The most straightforward approach to realizing a single-frequency, single-mode semiconductor laser is to use a Fabry-Perot resonator built with two mirrors with different reflection coefficients, coated at the end facets of the lasing structure, to built a Fabry-Perot (FP) laser.

The distributed-feedback Bragg (DFB) grating laser achieves its single wavelength operation using a distributed Bragg grating structure extended over the entire junction. These lasers have a very narrow emission line, are simpler to fabricate than other single-frequency semiconductor lasers, and therefore are widely used as a single-wavelength generator. Distributed Bragg reflector (DBR) lasers have two Bragg grating zones working as distributed mirrors forming a resonator for the active medium contained between these zones.

Semiconductor tunable lasers (STL) have a similar structure as DBR lasers. There are two sections between the Bragg reflectors: an optical-gain section, similar to the optical-gain section of DBR lasers, and a tuning section, which changes the refractive index of the active layer underneath the contact layer of this section. The changes of the refractive index of the active layer produced in the tuning section changes the wavelength of the resonator made between the two Bragg grating reflectors, thus achieving the laser tuning to a particular wavelength.

Receivers
An optical receiver consists of an InGaAs photodiode connected to a pre-amplifier, usually fabricated with GaAs technology. InGaAs is the semiconductor material of choice for photodiodes used in optical communications, due to its spectral characteristic. The response of InGaAs matches the wavelength range from 1280 nm up to 1650 nm used in optical communications. The photodiode has a bias voltage and the amplifier may have an offset control and, additionally, a gain-control input. The photodiode and the amplifier can be either in different packages or they can be mounted in the same package.

Optical amplifiers have two great features for optical communications. They amplify the optical signals using only optical effects, and they exhibit full scalability in terms of communication speed or bit rate, or of the communication protocol.

Erbium-Doped Fiber Amplifiers
Erbium-doped fiber amplifiers (EDFA) use the quantum effect of stimulated emission into a three-level quantum system to amplify optical signals. In one quantum mechanism, pump radiation with wavelength of 980 nm brings the Erbium atoms from the ground energy band A to an intermediate energy band C. The atoms will stay on band C about 10 µs, which is the lifetime of band C, and will leave this band through a non-radiative transition to the metastable band B, the upper level of the amplifying transition. Into another quantum mechanism, a pump wavelength of 1480 nm brings the Erbium atoms from the ground state A to the upper energy levels of the band B. The lifetime of Level B is quite long, about 10ms, which allows many Erbium atoms to be pumped into this level. The Erbium atoms leave level B by two quantum mechanisms:

• Transitions stimulated by an optical signal in the wavelength range 1500 nm to 1580 nm that amplifies the input optical signal
• Spontaneous transitions that generate optical noise.

One of the main drawbacks of EDFA is nonlinear gain in the optical region between 1530 nm and 1560 nm.

Raman Amplifiers
The Raman amplifier (RA) uses the stimulated Raman scattering inelastic-interaction described in the Chromatic Dispersion section, to transfer the energy from the pump wave to the signal wave. The Raman pump wave with about 1W power or more co-propagates with the signal and is sent to the monomode fiber carrying the signal, using a coupler and an optical isolator. Some of the advantages of Raman amplifiers are that the amplification of the signal is done in the same single-mode fiber used to carry the signal, and that signal amplification can take place in the L band, beyond the spectral limits of Erbium-doped amplifiers.

There are also some drawbacks of Raman amplifiers. The high power levels necessary to produce SRS brings the core into a non-linear region, which can generate four-wave mixing and other unwanted non-linear effects. In addition, there is an increase of optical crosstalk from lower-to-higher channels.

DWDM Spectrum
Optical carriers are usually amplitude modulated and the modulation spectrum appears as side lobes of the carrier frequency. For reliable communication, it is obvious that the frequency gap between the left and right side lobes of adjacent channels must have a small value to allow a maximum number of channels into a certain wavelength range. This frequency gap must also be large enough to prevent the eventual overlap between the neighboring side lobes. You cannot simultaneously increase modulation speed and decrease the channel spacing without introducing crosstalk between the channels. In addition, you must consider the non-linearity of the fiber when the total optical power goes beyond a certain level. There must always be a balance between channel spacing, transmission speed per channel, and total power launched into the fiber.

Wavelength Multiplexing / De-Multiplexing
There are two approaches to implement DWDM devices such as wavelength multiplexers, de-multiplexers, and add-drop filters:

• The fixed-frequency approach widely used to date is to assign a wavelength for each device input and, eventually, for each output as well. This is an advantage since the device is quite simple and does not need control for its operation. Its relative simplicity is reflected in low cost.
• The tunable approach of these devices has the great advantage of optimizing the transfer characteristic of the device to any wavelength. Tunability also offers tracking capabilities to match device characteristics with the signal, even for small departures of the signal from the ITU grid. In tracking mode, the device can provide to a network controller information on the quality of the optical signal. Moreover, the tunable device control allows direct interfacing of the device with the optical network monitor.

Optical Networks
If the end points of a linear communication link are coincident, the link becomes circular, as in Synchronous Optical Network (SONET) architecture, known also as Synchronous Digital Hierarchy (SDH) architecture in ITU terminology. Users connect to SONET through Add-Drop modules. The different rings of a SONET network are connected through ring interconnect elements. If many linear communication links are connected into a web configuration, the outcome is an optical network with a mesh architecture, having some nodes A, B, C, D, and E. Every node of the mesh network has a cross-point switch to realize the connections between different links (Figure 1).

Figure 1:  Mesh optical network architecture

Features of Optical Switches
Optical switches are essential elements of an optical network. You use the switches to either route optical signals between different optical fibers in cross-point switches of a mesh configuration or to route the signals in the same fiber of a lambda-switching network when the switch is combined with optical filters. A very useful device for a lambda-switching network would be a combination between a tunable optical filter and an optical switch, called a tunable switch. This type of device can enable or disable a specific wavelength signal into a wavelength router.

An optical switch can work for any wavelength used in optical communications, at any bit rate, with any communication protocol, and does not require optical-electrical-optical conversion to realize its function. The optical switch is also bi-directional, any port having the possibility to work either as an input or output. Actual optical switches have a switching time of about 10 milliseconds.

Configuration of Optical Switches
A few basic formats of optical switches are:

• 1x2 format has three optical ports: A, B, and C. It switches the optical signal between port A and ports B or C, depending on the control signal applied to it.
• 2x2 format has four optical ports: A, B, C, and D. Depending on the control signal, the switch can be in the bar state, when it connects ports A to C and B to D, or in cross-state, when it connects ports A to D and B to C.
• MxM format is a matrix switch that can realize all possible connections between the ports X1, X2, …Xm and the ports Y1, Y2, …Ym, as specified by the control command that is now more complicated than in the prior two cases.

The optical-communications market is definitely moving towards an all-optical structure, with mesh architecture and wavelength routing. Tunable lasers and tunable optical filters are some of the key elements to realize these goals. Other key issues to enable long-haul high-speed communications are polarization and optical switches.