Cart 0

Single-Mode Fibers for Communications

This is a continuation from the previous tutorial - Glan-type prisms.


1. Introduction

System performance can be maximized and total system cost savings can be realized by choosing an optical fiber design optimized for a particular system application.

The cabled optical fiber that forms the backbone of the physical layer is one part of an optical transmission line that also comprises amplifiers and dispersion compensation modules (DCMs).

The designs of the amplifier, DCM, and cabled transmission fiber are not mutually independent, and an integrated view of the transmission line design is necessary to optimize performance and drive cost out of the total system.

A digital transmission system relies on the ability of a receiver to discriminate whether a transmitted bit is a ‘‘1’’ or ‘‘0.’’ The ability to do this can be degraded by multiple factors.

First, the signal-to-noise ratio (SNR) can be degraded in either the optical or the electrical domain. Insufficient optical SNR (OSNR) can be due to noise in the transmitter, optical attenuation in the fiber span, the injection of noise by an amplifier, or insufficient receiver sensitivity.

Second, intersymbol interference (ISI) occurs as adjacent bits (symbols) spread into one another and overlap due to dispersion in the fiber. ISI can cause a ‘‘0’’ to be corrupted by a neighboring ‘‘1.’’ ISI arises from both chromatic and polarization mode dispersion. In a purely linear system, ISI can be fully compensated, at least theoretically, by a dispersion compensation device.

Nonlinearity in the optical transmission line can permanently distort the shape of an optical pulse. The combination of nonlinearity with dispersion leads to uncorrectable ISI penalties. Impairments can limit the reach, bit rate, or capacity of a system, whereas devices that mitigate impairments add cost and complexity to a system.

In choosing optical fiber design targets to maximize performance in specific application segments such as metro, long-haul, or ultralong haul, a clear understanding of how fiber properties affect current and future system performance must be combined with an understanding of the system cost impact.

The technologies used to amplify signals, compensate dispersion, and route wavelengths will interact with the fiber properties to affect the cost–performance balance. The constraints due to practical considerations of cabling (microbending and macrobending sensitivity), limits on cutoff wavelength, the availability of ultrapure materials, and cost of fabrication are also important in developing a new
commercial product.

This tutorial explores several key network segments for which optimized optical fibers have been designed and commercialized. This tutorial focuses on the application-specific transmission fibers that are deployed in optical cables in the outside plant as the primary infrastructure of a communications system.

Fibers for dispersion compensation or nonlinear signal processing are deployed in modules in a terminal or hut and are the subject of other tutorials.

First, an overview of the system limitations and penalties that drive optical fiber design is presented. A summary of International Telecommunication Union (ITU) fiber standards follows.

Fibers optimized for lowest loss in shorter reach applications are discussed next. Then we present design principles for optimizing an optical transmission line for high bit rate, high-capacity dense wavelength diversion multiplexing (DWDM) applications.

Finally, the waveguide design and properties of nonzero dispersion-shifted fibers (NZDFs) are discussed in detail.


2. System Impairments Influencing Fiber Design


2.1 Limitations from Optical Signal-to-Noise Ratio

Two cases are considered in which optical transmission may be limited by the degradation of OSNR. The trivial example is a low-cost nonamplified link that is simply loss limited.

A receiver has a specified sensitivity in dBm (logarithmic unit of power), and the power budget specifies the maximum loss allowed to achieve the desired bit error ratio (BER) at the data rate.

If we assume a 10-Gbps PIN diode receiver sensitivity of \(-18\) dBm and a launch power in the range of a few dBm, then a loss budget of approximately 20 dB is available for fiber, splice, and connector loss.

Carriers often budget for a worst-case cabled fiber loss of approximately 0.25 dB/km at 1550 nm. As an example, allowing 2 dB for splices and connectors results in a calculated span length of 72 km.

For many years, almost a third of the optical transmission window between 1260 and 1625 nm was unusable because of the presence of a large ‘‘water peak’’ at 1383 nm with loss as high as 2 dB/km. Today, zero water peak (ZWP) fibers, described in detail later, open previously unusable spectrum for low-cost coarse wavelength division multiplexing (CWDM).

The more interesting case is that of multiple spans, in which noise accumulates in an amplifier chain. An expression that approximates the OSNR for a link comprising \(N_{span}\) spans of loss \(L_{span}\) in which each amplifier has noise figure, \(NF\), is given as


where all units are in dB or dBm as appropriate and \(P_{ch}\) is the span launch power.

The required OSNR scales inversely as the data rate increases, increasing by 6 dB from 10 to 40 Gbps. This is true because for a fixed received power, the number of photons per bit decreases by the same factor that the bit rate increases. All else being equal, a requirement of 6 dB higher OSNR would decrease the reach of a 40-Gbps system by a factor of four relative to a 10-Gbps link.

Equation (5.1) shows that the most effective method for increasing OSNR in an amplified span is to reduce the loss by spacing amplifiers more closely, because the OSNR improves linearly with span loss but only rises logarithmically with the number of amplifiers. However, this option is available only to the submarine system designer, in which repeater spacing is flexible.

Distributed Raman amplification is a key-enabling technology for 40-Gb transmission, improving OSNR by about 3 dB, as well as forward error correction and advanced modulation formats. Transmission fiber designs that improve the efficiency and cost-effectiveness of Raman amplification are described in the following subsections.


2.2 Limitations from Intersymbol Interference

ISI occurs when pulses broaden out of their assigned bit slots because of some form of dispersion in system components. Chromatic dispersion arises from the wavelength dependence of the propagation constant in the fiber and is proportional to fiber length and laser line width.

The requirements on chromatic dispersion become very severe for higher bit rates, increasing as the square of bit rate. PMD arises from the small usually randomized birefringence of the fiber and grows with the square root of fiber length.

Systems operating at 2.5 Gbps or less permit the use of low-cost directly modulated lasers with wider line widths arising from chirp. Systems running standard NRZ modulation at 10 Gbps over more than 10–20 km of G.652 standard single-mode fiber (SSMF) rely on externally modulated, CW sources with very narrow intrinsic line widths.

As in the case of OSNR, we may distinguish two cases of dispersion limitation: one applicable to low-cost systems, in which it is desirable to avoid in-line dispersion compensation altogether, and the other relevant to long-reach systems based on concatenated amplified spans.

The simplest ISI limit occurs for uncompensated transmission, in which the dispersion tolerance of a ‘‘receiver’’—often defined as that dispersion in ps/nm resulting in a 1-dB power penalty—limits the reach of link.

For SSMF, with a 1550-nm dispersion of approximately 17 ps/nm-km, that limit is usually reached at 60–80 km. At the time of this writing, efforts are underway to write standards for electronically equalized receivers that extend uncompensated reach with G.652 SSMF from 80 to 120 km. However, optical fiber designs with one-fourth the dispersion of SSMF can increase the reach of uncompensated links by a factor of four.

Maintaining tight dispersion tolerances over many channels in an amplified DWDM system comprising many spans once posed a significant challenge for transmission line design.

This obstacle was overcome by the milestone development of the slope-matched DCM, which is capable of tightly compensating dispersion across the entire C- or L-band in one broadband device.

For the purpose of this chapter, it is critical to note that transmission fibers can be designed so they co-optimize the design of compensating fiber in the DCM for optimum broadband compensation.

PMD is a special challenge because it is an inherently statistical quantity that depends sensitively on the mechanical state of the fiber. The single number known as the PMD coefficient used to characterize an optical fiber is intended to describe the (normally) maxwellian distribution of differential group delay (DGD) values that a fiber can assume.

A single fiber displays its full range of DGD values as its mechanical stress state changes through all possible configurations, usually varying with temperature cycles or mechanical vibrations (e.g., for a cable laid alongside a railroad track).

The link design value (LDV) is used to identify the worst-case PMD that a link comprising many cabled fiber segments will experience. As shown later in this tutorial, there is a small but nonzero probability that DGD in a system will fluctuate through very high values that can halt transmission for short periods each year.


2.3  Limitations from Nonlinearity

Nonlinear impairments can broaden the frequency content and temporal evolution of a single pulse, cause crosstalk between pulses in adjacent channels, and cause two adjacent channels to produce noise in a third channel.

Nonlinear impairments are most pronounced in very long systems in which small nonlinear products are able to accumulate. Cross-phase modulation (XPM), as an interchannel impairment, is usually the dominant nonlinearity in 10-Gbps DWDM systems.

Four-wave mixing(FWM)can be essentially eliminated as an impairment between channels in 10-Gbps systems as long as the fiber dispersion across the band is greater than approximately 2 ps/nm-km.

Self-phase modulation (SPM), XPM, or FWM between temporally adjacent bits at the same wavelength (i.e., intra-channel nonlinearities) tend to dominate in 40-Gbps systems.

The OSNR improves linearly (on a decibel scale) as transmitted power increases; however, the penalty due to nonlinear impairments increases with transmitted power.

An optical transmission system is often designed to operate in the sweet spot that maximizes OSNR but avoids significant nonlinear distortions. It will be shown that the same transmission fiber design strategy that allows co-optimization of broadband dispersion compensation also reduces system nonlinear penalties.


2.4 Limitations from Amplifier Technology

The advent of the erbium-doped fiber amplifier (EDFA) in the early 1990s revolutionized the economics of optical transmission in several ways. Not only did it make expensive optical-to-electronic-to-optical (OEO) conversion, also known as regeneration, necessary only at the endpoint of a link, but it also favored adding as many WDM wavelengths as practical within the amplifier bandwidth.

The EDFA is most efficient in the C-band but has been successfully extended as a practical L-band amplifier as well. Although the C- and L-bands together provide tremendous capacity with DWDM technology at 40 Gbps, it is prudent to consider future expansion needs because optical fiber cable is rated for a lifetime more than 20 years.

Raman amplification not only is an enabling technology for 40 Gbps but can also provide gain where other technologies are not available. The gain spectrum for stimulated Raman scattering is based on the phonon (vibrational) spectrum of the silica glass, not on the electronic spectrum of dopants in the glass.

Essentially the peak of the Raman gain curve will lie approximately 13 THz lower than the pump frequency, which is approximately 100 nm to the red of the pump wavelength in the telecommunications bands. This means, for example, that Raman pumps can be placed at 1410 nm to provide gain for S-band channels near 1505 nm.


2.5 Can Fiber Design Be Used to Optimize a Transmission System?

The question addressed in this tutorial is how can optical fiber design facilitate the mitigation of these impairments to enable higher capacity, more cost-effective transmission? To answer that question, one must address three issues:

  1. The impact of fiber of properties on transmission impairments
  2. The fiber properties that can be manipulated
  3. The design tradeoffs that must be considered

The first question regarding how fiber properties affect transmission impairments is more complex than commonly supposed. The cabled fiber is often referred to as the ‘‘transmission fiber’’ (Tx fiber) to distinguish it from the dispersion compensation fiber (DCF).

The Tx fiber is only one part of a transmission line, but the Tx fiber properties critically affect the design and performance of the other components of the transmission line. It is the system OSNR, PMD, and nonlinearity that must be optimized, not the properties of the individual components.

Figure 5.1 illustrates the basic elements of the optical transmission line.


Figure 5.1 The basic elements of an optical transmission line, the designs of which are highly interdependent.


The loss of the fiber and connectors must be offset by either EDFA, Raman amplification, or a combination of both. For distances greater than approximately 80 km, the chromatic dispersion of the fiber must be compensated to eliminate ISI.

For older fibers, or even modern fibers manufactured without ‘‘spinning,’’ some form of PMD compensation may be required. The key point here is that the optical properties of the Tx fiber directly impact the requirements for efficiency, linearity, loss, and PMD of the other elements of the transmission line. 

The second issue concerns which fiber properties can be effectively manipulated by the fiber design.

The primary fiber property available for modification is the dispersion curve, which describes how an optical pulse of a given wavelength and spectral width will broaden in time as it propagates down the fiber. Altering the dispersion curve necessarily involves altering the effective area (\(A_\text{eff}\)) of the fiber as well, which also affects fiber cutoff wavelength.

Finally, it is critical to understand that design tradeoffs are inevitable.

The Maxwell equations dictate that key transmission properties of an optical fiber cannot be varied independently. In general, adding waveguide dispersion to the fiber to modify the dispersion curve will reduce the mode field diameter (MFD) and \(A_\text{eff}\).

System design issues dictate that the ideal fiber will have a lower dispersion at 1550 nm than standard fiber, as well as a lower dispersion slope. The cartoon in Fig. 5.2 illustrates the fundamental problem.


Figure 5.2  The Maxwell equations constrain the design of nonzero dispersion-shifted optical fiber so that tradeoffs must be made. It is not possible to independently choose the dispersion slope, bending losses, and effective area of a fiber, unless the fiber cutoff is allowed to rise above the operating wavelength. Bending losses are fixed by restrictions on allowed cable and handling losses. It is generally not possible to retain a desirable low slope while retaining a large effective


If we assume that the cabled cutoff must be maintained below 1260 nm to permit 1310-nm applications, then manipulation of the dispersion curve to reduce the dispersion and dispersion slope must be balanced against the need to maintain a reasonable \(A_\text{eff}\) while keeping bending losses low.

\(A_\text{eff}\) must be kept moderate to high to limit nonlinearities. Low macrobending and microbending losses are critical for good cable performance in the field.

Once we choose any two of the attributes in Fig. 5.2, the third is chosen for us by the Maxwell equations. Intelligent decisions on these tradeoffs are the key factor in fiber design and require close coupling of expertise in fiber design and processing with understanding of optical transmission systems.


3. Overview of ITU Standards Fiber Categories

The ITU is an agency of the United Nations and has, over the last 20 years or so, categorized single-mode fiber to assist suppliers and their customers in providing optical fiber designed to meet specific telecom applications.

The development of ITU recommendations is currently occurring in ITU-T Study Group 15 (ITU-T is the telecommunications standards branch of ITU). It is to be noted that many fiber properties are actually specified in cable, because that is generally the way the fiber is used, and many fiber properties can change once the fiber has been cabled.

Other statistically based properties are specified on a link basis. It should be emphasized that ITU standards are developed in a multiparty process that sometimes favors weakening requirements in order to gain broad consensus.

For example, it will noted that ITU requirements on PMD are relatively loose compared to the system requirements to be described in Section IV, Table 5.1.

Although ITU standards play a vital role in educating and informing the broader fiber optic community, they should be viewed as necessary but not sufficient conditions for an optical fiber to enable a particular application.

The most fundamental way to group optical fiber designs is by the characteristics of the dispersion curve. Figure 5.3 shows a selection of historically significant dispersion curves labeled by their ITU categories.

Characteristics of the dispersion curve that may be significant are the point where the dispersion is zero, called the zero-dispersion wavelength (ZDW) or \(\lambda_0\), the value of dispersion in the transmission band, and the slope of dispersion across a band. The significance of these quantities is described in detail in Sections IV and V.


Figure 5.3  Illustration of the International Telecommunication Union (ITU) fiber categories using chromatic dispersion curves representative of various fiber designs. The chromatic dispersion properties of fibers manufactured according to a particular fiber design will vary within stated tolerances determined by the design sensitivity and manufacturing process limits.


The original standard single-mode fiber is specified in ITU recommendation G.652. This fiber has a MFD in the range 8.6–9.5 microns, a maximum cable cutoff wavelength of 1260 nm and ZDW in the range 1300–1324 nm. This fiber typically has a chromatic dispersion of 17 ps/nm-km at 1550 nm (which can be excessive for dispersion sensitive applications).

Although there are several categories within G.652, the most modern is G.652D fiber that has low attenuation in cable (maximum of 0.3 dB/km at 1550 nm), good PMD performance (better than 0:2ps/km1/2 link design value), and low water peak (LWP) attenuation at 1383 nm.

The LWP designation requires that loss at 1383 nm be less than the maximum loss at other wavelengths from 1310 to 1625 nm. Based on 1310-nm specifications, LWP generally means 1383 nm loss <0.35 dB/km.

However, some manufacturers have virtually eliminated the water peak by careful processing. Because the E-band is available on account of the LWP loss, the G.652D fiber is optimized for full spectrum use at data rates up to 10 Gbps (STM-64). It is important to note that some manufacturers routinely offer G.652 fibers with a PMD specification three to four times tighter than required by the standard. Low PMD is critical for future upgrades to a fiber network.

The dispersion-shifted fiber (DSF) was developed in the late 1980s to support transmission in the low-loss 1550-nm window. These systems transmitted a single channel in the vicinity of the loss minimum at 1550 nm where the ZDW was located.

This allowed increasing the loss limited transmission distance. DSF was standardized in G.653 with a ZDW in the range 1500–1600 nm. Today, this fiber has limited application because of nonlinearities that occur between WDM optical channels when they are close to the ZDW. The embedded base of DSF in some countries has been upgraded by moving to L-band systems for DWDM.

Cutoff-shifted fiber was standardized in G.654 to provide lower loss and allow higher optical power for transmission over long distances. This fiber has a larger MFD than G.652 (some categories go as high as 13 micron), a cutoff wavelength as high as 1530 nm, with low attenuation limits (0.22 dB/km at 1550 nm) and a tight PMD specification (as low as 0:2ps/km1/2).

The chromatic dispersion is specified at 1550 nm and is similar in size to that of G.652 fiber. This fiber has been applied in submarine systems, in combination with a cabled inverse dispersion fiber, as well as long unamplified links.

As high bit rate transmission systems were designed using wideband EDFAs and the low fiber loss in the C-band, it became clear that the high chromatic dispersion of G.652 fiber at 1550 nm would limit transmission capacity because of dispersion-related signal impairment, and that the zero dispersion of DSF near 1550 nm would result in signal impairment related to nonlinear propagation effects. A better solution was found by shifting the ZDW away from the C-band.

A balance of fiber properties for long-haul applications is found in the NZDFs that are standardized in G.655. These fibers support transmission rates of 40 Gbps (STM-256) over long distances. The key features are low (but nonzero) chromatic dispersion in the C-band and low PMD (<0.2ps/km1/2).

Cable cutoff is held below 1450 nm. A classification is now being developed in ITU that will group G.655 fibers according to ‘‘low’’ or ‘‘medium’’ dispersion. In addition, limitations to the dispersion over the entire 1460- to 1625-nm range will be specified.

The G.655 upper limit on PMD should be contrasted with the tighter and more realistic requirements discussed in Section IV, because it could be argued that PMD is the most critical specification for upgradeability to 40-Gbps operation.

A final fiber category is the CWDM/DWDM optimized fibers outlined in G.656. The original purpose of this category was to have low dispersion from 1460 to 1625 nm to decrease ISI that limits uncompensated CWDM transmission.

However, the requirements evolved substantially during consideration and debate. In the final analysis, G.656 can be considered to be the wideband Raman-enabled fiber standard. These fibers have specified performance over the 1460- to 1625-nm wavelength range (similar to the newer G.655 table).

In fact, fibers in the proposed medium dispersion category of G.655 also fulfill the requirements for G.656. The key feature of the G.656 fiber is that there is a minimum chromatic dispersion of 2 ps/nm-km at 1460 nm, which enables good performance (free from nonlinearities) for signal channels, as well as Raman pumps at short wavelengths.

Some G.655 fibers are lacking in this regard. Fibers according to G.656 fibers allow highest performance with optical channels spaced over a wide band at 40 Gbps and over long distances.


4. Optical Fibers for Reduced Attenuation

For systems in which span loss limits performance in unamplified systems, the fiber loss (in cable) is the key enabling performance parameter. Two categories of fiber for such applications have been commercially important: ZWP peak and pure silica core fibers.

In contrast to fibers with novel dispersion properties controlled by a special profile design, loss reduction is primarily accomplished by improving the materials chemistry and physics of the silica optical materials and carefully controlling fiber processing.

Since the beginning of silica optical fiber development, the (previously) ubiquitous loss peak centered around 1383 nm was recognized as the vibrational second-overtone absorption of the hydroxyl group OH.

This OH or ‘‘water’’ peak causes increased optical loss from about 1360 to 1460 nm, a wavelength region now designated as the E-band by standards bodies. Because of the high OH loss in typical commercial single-mode fibers, most telecommunication systems until recently avoided the E-band and instead used the two windows at either side of the OH peak, namely, the O- and C-bands centered at 1310 and 1550 nm, respectively.

Although the benefits of completely eliminating the OH loss in commercial fibers had been obvious for more than 2 decades, the enormous technical and commercial challenges of realizing a viable zero-OH manufacturing fostered a belief in its impracticality. Today ZWP performance (\(L_\text{1383nm}\) < 0.31 dB/km) can be guaranteed at essentially no additional cost to a carrier.


4.1  Pure Silica Core Fiber

The design of pure silica core fiber seeks to minimize Rayleigh scattering losses due to nanoscopic fluctuations in the index of refraction (determined by either the glass density or the chemical composition).

Average loss of 0.168 dB/km is commercially obtained at 1550 nm, while the loss minimum near 1580 nm is a few thousandths lower still. This is achieved by eliminating germania as the dopant to create the waveguide.

Instead, the cladding is heavily fluorinated to reduce its index of refraction, creating a profile very similar to the matched or depressed cladding profiles, with the exception that all index of refraction values are shifted down by approximately 0.35%.

However, the properties of silica core fiber—and the index profile itself—are extremely sensitive to the drawing conditions. Because the high-temperature viscosity of pure silica is much higher than that of doped silica (whether doped with Ge or F), most of the tension during draw is carried by the core.

In the extreme case, the stresses and strains that remain in the final drawn fiber can alter the refractive index profile to the degree that the core index is reduced close to that of the F-doped cladding.

Furthermore, the high viscosity of pure silica leads to a relatively more rapid quenching of the glass nanostructure as the fiber rapidly cools. This results in less time for annealing of density fluctuations and thus greater Rayleigh scattering. As a result, silica core fiber is typically drawn at speeds of 1–2 m/sec to achieve the loss performance noted earlier.

This slow draw speed, combined with a complex preform fabrication process, makes silica core fiber traditionally very expensive compared to the highest quality G.652D fiber.

As a result, the application of silica core fiber has been limited primarily to unamplified (or ‘‘unrepeatered’’) submarine systems of several hundred kilometers in length, where the lower loss is a requirement for system operation.


4.2 Zero Water Peak Fiber

A more practical and commercially viable low-loss fiber design has been one that eliminates the water peak loss from standard single-mode fiber, whether of matched or depressed cladding profile design.

Such fiber meets and exceeds the ITU G.652D standard and has become the de facto standard for high-quality fiber in many applications spaces. The water peak loss at 1383 nm has historically raised loss across approximately 30% of the available spectrum in the lowloss window between 1260 and 1625 nm.

For many years, the water peak specification limit in standard fibers was as high as 2 dB/km for several major fiber suppliers. The elimination of water peak loss opened up spectrum for application of low-cost 16-channel CWDM, as well as removed a major limitation to the efficiency of Raman amplifiers.

It was in fact the realization of such zero water peak (ZWP) fiber or, in less demanding applications, a LWP fiber, that provided the key enabling component and the impetus for the serious work on CWDM components.

In the medium- and short-distance local and access networks, low-cost CWDM with ZWP/LWP single-mode fiber is ideally suited to aggregate traffic from the rapidly growing number of FTTx broadband users and deliver it to the core network in a cost-effective way.

It should be noted that although the focus of ZWP/LWP fiber development and applications has been primarily for G.652 fibers with a zero dispersion near 1310 nm, aspects of the zero-OH technology described here can also benefit fibers of other designs and for different applications

For example, LWP NZDFs have increased bandwidth for DWDM applications and improved Raman pumping efficiency at 1450 nm because of the lower fiber loss at that wavelength.

Water peak reduction is now a common feature of many fibers, including G.655 and G.656 NZDF fibers. This section focuses on the requirements necessary to manufacture and specify ZWP G.652 fibers of the basic profile shapes, whether of the matched or depressed clad profile design.

While it has always been clear that to achieve LWP or ZWP fibers, one has to keep the OH contamination in silica to an extremely low level (i.e., to <0.1 ppb of OH in the core of ZWP fiber for <0.005 dB/km of added loss at 1383 nm, for instance), the biggest challenge was to devise a practical, high-yield, and low-cost manufacturing process.

To put this challenge in perspective, commercial single-mode fibers are still sold with a 1383-nm loss as high as 2 dB/km; this value is equivalent to about 40 ppb of OH in the fiber core. An optical preform comprises a core made from high-quality deposited glass, lower purity overcladding glass, and an interface between them.

The water peak loss at 1383 nm can be broken down into components as


where \(L_\text{Rayleigh}\) represents the background Rayleigh scattering level at 1383 nm, which is independent of OH contamination. This value is approximately 0.26 dB/km. \(L_\text{deposit}\) represents the contribution from light propagating in the deposited core material, which is the highest purity glass, where the optical power is maximum. This glass may be formed with the VAD, MCVD, OVD, or PCVD methods, all of which have shown the capacity to support the LWP performance level (typically \(L_\text{total}\) < 0.35 dB/km for G.652C/D fibers or \(L_\text{total}\) < 0.38 dB/km for G.655 or G.656 fibers).

However, the VAD core method has proven most adept at making ZWP fiber in large preforms. The VAD or OVD methods allow for an explicit dehydration step during glass deposition where silica in a porous soot form can be exposed to Cl2 at temperatures above 800°C to eliminate OH before sintering.

VAD cores are formed as solid rods, while OVD soot is deposited onto a mandrel and requires an additional step, to collapse the glass annulus after sintering, during which ultimate dryness must be maintained to avoid contaminating the centerline of the preform with OH.

The PCVD and MCVD processes deposit glass inside a substrate tube, whose additional contribution to 1383-nm loss is represented above by \(L_\text{tube}\), referring to Fig. 5.4.

Even a high-quality substrate tube can have OH concentration as high as 200 ppb, although higher purity is available at a cost premium. Substrate tube glass is approximately 9–10 microns from the fiber centerline for MCVD or PCVD fiber drawn from large-diameter preforms, so the optical power is lower.

Nevertheless, \(L_\text{tube}\) can be on the order of 0.05 dB/km. Like OVD, PCVD and MCVD also form an annular glass body requiring a very high temperature collapse process during which the centerline of the preform must be protected from OH contamination.

Neither PCVD nor MCVD allows for a separate dehydration step. The PCVD process is noted for its ability to incorporate a wide range of feedstock materials, including any impurities.


Figure 5.4  Cross-section, not to scale, of an optical fiber preform. Left figure is appropriate for a monolithic core rod made with OVD or VAD. A typical D/d ratio would be in the range of 3–5. Right figure shows the case in which core material is deposited inside a substrate tube in the MCVD or PCVD processes. In this case, D/d may be in the range of 2.5 or less.


The contribution to OH loss from the interface between the higher purity glass core rod (regardless of core fabrication technology) and the generally less expensive cladding glass is represented by \(L_\text{interface}\).

For rod-in-tube technology, this interface is formed by collapsing one glass surface onto another at high temperature. The opportunity is thus present to contaminate the surfaces and diffuse moisture into the core.

In the case of soot overcladding, this interface is the result of depositing high-density soot on glass, yielding an interface that is more difficult to dehydrate than bulk soot.

In the case of plasma overcladding, molten quartz particles are sprayed onto a glass core. \(L_\text{cladding}\) is generally negligible because the total power in the cladding is relatively small, on the order of 0.01–0.1% depending on fiber design and D/d.

It is important to note that it is always possible to minimize \(L_\text{interface}\) and \(L_\text{cladding}\) by raising the ratio D/d, shown in Fig. 5.4, where D is the diameter of the core body (before overcladding) and d is the diameter of the GeO2-doped region of the waveguide.

A large D/d ratio means a higher percentage of the total glass is formed in the high-purity core deposition step, pushing the interface and the cladding glass farther away from the fiber centerline where the optical power is maximum.

Applying this tactic to make ZWP or LWP fibers generally requires making a smaller less economical preform. This is necessarily true since practical issues of fluid flow, heat transfer, or even legacy machine design limit the total amount of core deposit possible. Clearly smaller preforms or large D/d cores are only cost effective for more expensive specialty fiber products.

However, eliminating OH contamination is only part of the challenge. As a result of the manufacture and study of ZWP fiber, an important new hydrogen aging loss mechanism was discovered where the OH peak in ZWP/LWP fibers can grow by as much as a few tenths of dB/km or more when certain reactive atomic defects were present in the fibers to rapidly react with a trace amount of molecular hydrogen at ambient temperature.

Because of the ubiquitous nature of these reactive defects in silica fibers and the inevitable presence of small amounts of molecular hydrogen in fiber installations, a solution to the hydrogen aging loss problem must be found for ZWP/LWP fibers so the loss, particularly at the OH peak, remains permanently low throughout their service lifetime.

Thus, the following sections also include a discussion on the solution to the hydrogen aging loss problem and the development of the standards for hydrogen aging test.

Although the elimination or reduction of the OH loss is the key to ZWP/LWP fibers and low-cost CWDM applications, other fiber performance parameters are also important.

In particular, to support future migration to 40 Gbps and use of the L-band, high-performance in fiber PMD and macrobending, respectively, are required. But perhaps even more important than technical performance, the cost for manufacturing ZWP/LWP fibers must be low because it is the critical driver for any mass fiber deployment toward the end-users.

These different aspects of ZWP/LWP fiber performance, manufacturing process, and cost are discussed in the following subsections.

4.2.1  A Cost-Effective Fabrication Technology for ZWP Fiber

ZWP fiber was first developed by Lucent Technologies, Bell Laboratories in 1997 as the commercial AllWave ZWP fiber using the Rod-in-Tube (RIT) process.

A method for solving the technical and economic challenges of fabricating ZWP fiber is the Rod-in-Cylinder (RIC) and Overclad-during-Draw (ODD) manufacturing process. This fabrication process increases the preform size to more than 5000 fiber km and achieves low cost and ZWP fiber performance exceeding ITU G.652D standards.

Essentially, the RIC-ODD process entails a totally mechanical RIC assembly and drawing this large assembly directly into fiber with the ODD process. The RIC assembly is constructed by inserting a VAD core rod and an optional thin-walled first overclad tube into a large, hollow overclad cylinder of up to 170 mm outside diameter (OD), 60-mm inside diameter (ID) and 3 m in length to form a preform capable of yielding more than 5000 km of fiber.

The bottom end of the cylinder is machined into a conical taper and a hole is drilled through its walls so that a quartz plug-and-pin assembly can be inserted. The plug-and-pin is used to hold up the VAD core rod and the first overclad tube inside the cylinder.

The conical taper of cylinder bottom facilitates the initial seal and glass drop during the RIC-ODD draw process. The cylinder also has a quartz handle attached to the top end for handling by robotic manipulator.

When this completely mechanical RIC assembly is lowered into the draw furnace and a vacuum is applied through the hollow handle at the top, the overclad cylinder and the first overclad tube collapse onto the core rod and form a seal at the tapered end of the cylinder.

The plug-and-pin and rest of cylinder taper will then be melted and dropped off to begin the high-speed fiber draw as the RIC assembly is further lowered into the furnace and the vacuum-assisted ODD collapse continues. 

4.2.2 Maintaining ZWP Fiber Performance

To achieve ZWP performance, the VAD core soot body is dehydrated with chlorine prior to consolidation into a core rod. The VAD core rod is then etched with a hydrogen-free (i.e., ‘‘dry’’) plasma torch to remove its surface OH.

The first overclad tube and the overclad cylinder, both made by the OVD process with typically 300 ppb OH or less, are also lightly etched by acids to remove surface contaminants and OH.

From FTIR and fiber loss analyses, it has been found that less than 2 ppm OH remained at the interfaces (which had a width of about 1μm in a 125-μm OD fiber).

Such a small amount of interface OH will not adversely affect the fiber 1383-nm loss to a significant degree as long as the interfaces are placed at a sufficiently large distance from the fiber core (e.g., a clad-to-core ratio D/d > 3).

The median 1383- and 1550-nm losses for fiber produced by the RIC-ODD process achieved world-class levels of 0.276 and 0.187 dB/km, respectively. The essentially zero-OH loss (i.e., no observable OH peak) at 1383 nm in particular far exceeds the requirements for G.652C/D fibers.

It should be noted that a process like RIC-ODD has some inherent advantages for low 1383-nm loss:

  1. The VAD core bodies have no open holes or center lines like those appearing in the MCVD, PCVD, and OVD processes so it is relatively easy to keep OH completely out of core bodies with a separate dehydration and consolidation process
  2. The RIT or RIC overclad material, made by a separate OVD process, can also be completely dehydrated
  3. The surface OH on the core rods and overclad tubes or cylinders can be easily removed by plasma or acid etch
  4. It is fairly easy to keep the interfaces in RIT or RIC dry with a vacuum ODD process or with a more aggressive dehydration or etching procedure utilizing Cl- or F-containing gases during overclad.

In contrast, the complete dehydration of soot-on-glass interfaces for the alternative OVD and VAD overclad processes can be difficult, especially for large preforms.

For MCVD and PCVD processes where the glass layers are typically deposited without forming soot layers first, a complete dehydration of consolidated glass layers can also be difficult.

However, even in these cases, it is possible to greatly reduce the 1383-nm OH loss peak to the LWP level through vigilant practice of eliminating any leakage and avoiding any moisture contamination throughout the chemical delivery systems.

For ultralow fiber PMD, the patented draw process must be implemented to impart frozen-in spins in the core of the fiber. The result is a typical low-mode-coupled PMD \(\le0.02\text{ ps}/\sqrt{\text{km}}\) for RIC-ODD fiber, which again far surpasses the G652C/D requirements.

Cylinder surface preparation and cleaning procedures result in no interface problems such as airlines or bubbles and achieve a fiber break rate of less than 5/fMm at 100 kpsi (0.7 GPa) proof test.

The RIC-ODD fiber performance in core eccentricity, clad noncircularity, and fiber curl is excellent, supporting excellent typical splice loss of less than 0.02 dB. The G.652D depressed cladding fiber design can also be implemented with a similar large preform, ZWP, low PMD process to yield a bend-insensitive fiber well suited to FTTH access networks.

4.2.3 Hydrogen Aging Losses

The optical loss in fibers can degrade with time due to the chemical reaction between the inevitable atomic defects in fibers and the trace amount of molecular hydrogen normally present in or around optical cables.

This hydrogen aging loss is a particularly important problem for the ZWP/LWP fibers because all known hydrogen reactions with silica fibers will result in at least the loss increase at the 1383-nm OH peak (other additional hydrogen aging loss components are possible, depending on the defects and reaction types; see later discussion).

During the development of ZWP fiber, it was discovered that a very common but previously unknown silica defect is extremely reactive and can cause significant hydrogen aging loss at the OH peak upon brief exposure to trace amount of hydrogen, even at room temperature. We discuss this and two other types of hydrogen aging losses relevant to ZWP/LWP fibers, as well as the countermeasures against them.

Basically, there are three types of hydrogen aging losses that must be avoided in Ge-doped silica fibers to ensure reliability in optical transmission over the service lifetime of 25 years or more.

These hydrogen aging losses are caused by different types of atomic defects or impurities present in the silica fibers. The severity of hydrogen aging loss degradation is entirely dependent on the fiber manufacturing process and the purity of the silica material used.

The first two types of hydrogen aging losses involve two species of extremely reactive silica defects:

  1. A pair of nonbridging oxygen hole centers (\(\text{NBOHCs}\), \(\equiv\text{Si-O}\bullet\bullet\text{O-Si}\equiv\))
  2. The peroxy radical plus Si E' center (\(\equiv\text{Si-O-O}\bullet\bullet\text{Si}\equiv\)), where \(\bullet\) denotes an unpaired electron at the broken chemical bond.

These two types of silica defects (involving Si and O atoms only and no Ge) even at room temperature can react almost instantaneously with trace amounts of hydrogen and cause significant loss increases of up to a few tenths of dB/km or more.

The two hydrogen reaction mechanisms can be described as follows: 

\[\tag{5.3}\begin{align}\equiv\text{Si-O}\bullet\bullet\text{O-Si}\equiv+\text{H}_2\rightarrow\equiv\text{Si-O-H}+\text{H-O-Si}\equiv\\\text{NBOHCs}\qquad\qquad\qquad\qquad1383\text{ nm}\qquad\end{align}\]

\[\tag{5.4}\begin{align}\equiv\text{Si-O-O}\bullet\bullet\text{Si}\equiv+\text{H}_2\rightarrow\equiv\text{Si-O-O-H}+\text{H-Si}\equiv\rightarrow\equiv\text{Si-O-O-Si}\equiv+\text{H}_2\\\text{peroxy radical}+\text{Si E}'\qquad\qquad1383+1530\text{ nm}\qquad\qquad\qquad\qquad\end{align}\]

Hydrogen reaction (5.3) with NBOHCs results in an OH peak at 1383 nm. Hydrogen reaction (5.4) with peroxy radical and Si E' center results in an OH peak at 1383 nm and a ‘‘SiH’’ peak at 1530 nm, which are metastable and can decay at room temperature, although a significant fraction of the loss increases will remain after a few months.

Room-temperature hydrogen tests have shown that the two hydrogen reactions above typically reach saturation in less than 4 days in 0.01 atmospheres of hydrogen and this implies that the partial pressure of hydrogen in the cable installation needs to be much less than 4 ppm over the 25-year lifetime to avoid these two hydrogen aging losses: an impractical solution as the measured hydrogen partial pressure in cable installation is on the order of 400 ppm or more.

So, to reduce the risk of hydrogen aging loss, it is very important to minimize the above silica defects in the fiber manufacturing process. This can be achieved by adjusting the oxidation and reduction conditions in dehydration and consolidation in the preform making process as well as optimizing the fiber draw.

In addition, it is possible to passivate any remaining reactive silica defects in fiber by treating the drawn fiber spools with a small amount of deuterium at ambient temperature. Deuterium treatment can completely eliminate these particularly egregious hydrogen aging losses due to the two reactive silica defects.

The deuterium reactions work in a similar way as the hydrogen reactions (5.3) and (5.4):

\[\tag{5.5}\begin{align}\equiv\text{Si-O}\bullet\bullet\text{O-Si}\equiv+\text{D}_2\rightarrow\equiv\text{Si-O-D}+\text{D-O-Si}\equiv\\\text{NBOHCs}\qquad\qquad\qquad1900\text{ nm}\qquad\qquad\end{align}\]

\[\tag{5.6}\begin{align}\equiv\text{Si-O-O}\bullet\bullet\text{Si}\equiv+\text{D}_2\rightarrow\equiv\text{Si-O-O-D}+\text{D-Si}\equiv\rightarrow\equiv\text{Si-O-O-Si}\equiv+\text{D}_2\\\text{peroxy radical}+\text{Si E}'\qquad\qquad1900+2100\text{ nm}\qquad\qquad\qquad\qquad\end{align}\]

But the OD and SiD absorption losses are now harmless because they occur at much longer wavelengths (>1625 nm) and are completely outside the normal operating wavelength windows.

Furthermore, the reactive silica defects after being passivated by the deuterium reaction are no longer capable of causing additional hydrogen aging loss in the field.

Though less reactive, the third type of hydrogen aging loss that is of concern is when there is alkali (Na, Li, K, etc.) contamination in the Ge-doped silica fiber. Alkali contamination can be as low as a fraction of parts per million atomic (ppma) and still results in significant hydrogen aging loss over time.

This is because the activation energy for hydrogen reaction is greatly reduced when the normal high-activation energy Ge defects (which are inevitably generated in the fiber manufacturing process) interact with alkali impurities. The hydrogen reaction can be described as follows:

\[\tag{5.7}\begin{align}\text{Na}^+\qquad\qquad\qquad\qquad\text{Na}\qquad\text{Na}\qquad\\\equiv\text{Si-O}\bullet\bullet\text{O-Ge}\equiv+\text{H}_2\rightarrow\text{Si-O-H}+\text{H-O-Ge}\\\text{Na}^+\qquad\qquad1383\text{ nm}+\text{Long }\lambda\text{ loss}\end{align}\]

Alkali contamination can arise from the use of natural quartz material, insufficient purification, or contamination in preform processing. When there is alkali contamination in the hydrogen aging, loss has an OH peak as well as a ‘‘Long Wavelength Loss’’ that increases with wavelength beyond 1360 nm (Fig. 5.5).


Figure 5.5.  An example of short wavelength (SWL) hydrogen aging loss (left) and long wavelength (LWL) hydrogen aging loss (right).


We have developed an accurate quantitative model based on our extensive hydrogen studies and it predicts a hydrogen aging loss in the range of 0.02–0.04 dB/km from 1360 to 1625 nm (which includes the OH peak) for a Ge-doped fiber with 1 ppma alkali contamination after 25 years under typical cable operating conditions of 20°C and 400 ppm H2 (Fig. 5.6).


Figure 5.6.  Left: The 1550-nm hydrogen aging loss versus time is much higher for Fiber A with Na + Li = 1.35 ppma than that of Fiber B with Na + Li < 0.3 ppma in a 60 °C, 0.01 atm H2 test. The solid curves are from model calculations. Right: Model predicted hydrogen aging losses at 20 °C, 0.001 atm H2 service conditions again show the alkali contaminated Fiber A has a significant risk of loss degradation over its lifetime.


5. Optical Fiber Design Principles for Wideband and High Bit Rate Transmission

The design of transmission fibers to optimize performance of high-capacity, long-reach DWDM systems requires careful consideration of how fiber design affects the balance between OSNR, dispersion compensation, and management of nonlinearities across the entire system. Three critical aspects need to be considered.

First, precise dispersion compensation across the amplifier bandwidth is crucial to permit cost-effective 40-Gbps transmission. This requires excellent chromatic dispersion compensation and minimizing PMD.

Second, the contributions to nonlinearity from each element of the transmission line must be considered, and the nonlinearity of the composite transmission line should be minimized.

Third, Raman amplification is a key enabling technology for improving OSNR for high-bit-rate ultralong haul transmission.


5.1 Precise Dispersion Compensation

Pulse spreading due to the nonzero spectral width of the laser and dispersion of the Tx fiber must be compensated in systems longer than a few spans of SSMF (dispersion ~17 ps/nm-km at 1550 nm).

With the NRZ modulation format, the allowable dispersion (in ps/nm) corresponding to a 1-dB power penalty is approximately given by 100 Gbps-nm-1/B2, where B is the bit rate (in Gbps).

The squared dependence on bit rate is due to the fact that the transmitter line width is approximately equal to the modulation bandwidth, while the temporal width of the bit is inversely proportional to B.

Thus, a DWDM system running at 40 Gbps over the full C-band is limited to approximately 60 ps/nm of accumulated residual dispersion across all wavelengths from 1530 to 1565 nm, compared to a limit of approximately 1000 ps/nm for a 10-Gbps system.

The requirement of 60 ps/nm over 1000 km means that the transmission fiber and the DCF must be sufficiently matched to give a net 0.06 ps/nm-km residual dispersion when paired together.

Failure to meet this limit would necessitate the use of expensive per-channel compensation at the terminal. This section shows that the ability to design a wideband DCM to meet stricter limits on residual dispersion is closely related to the design of the Tx fiber.


5.2. Dispersion Compensation Fiber Technology

The development and commercialization of DCF in robust packaging has been one of the two most critical enabling technologies for the deployment of cost-effective DWDM transmission over amplified spans.

DCMs are passive fiber-based devices located at the amplifier, and often at terminals as well, that are capable of compensating the dispersion of the transmission fiber span over a range of wavelengths.

In addition to compensating dispersion itself, the periodic use of per-span, in-line dispersion compensation at amplifier sites is a key element in dispersion management techniques to control nonlinearities such as cross-phase modulation and other nonlinearities at 10 Gbps and beyond.

Key work on the design and fabrication of DCF with negative dispersion and negative dispersion slope began in the early 1990s. Successful large-scale manufacturing was in place by 1998.

Transmission fibers used in terrestrial applications have positive dispersion and positive dispersion slopes, as shown in Fig. 5.3.

DCFs have high negative dispersions in the range of -100 to -250 ps/nm-km, as well as negative slopes. To achieve the negative dispersion and dispersion slope, a large amount of waveguide dispersion must be added by strongly confining the mode in a narrow and high \(\Delta\) Ge-doped core, surrounded by a low \(\Delta\) deeply F-doped trench, with an additional ring of positive \(\Delta\) Ge-doped silica surrounding that.

DCF index profiles may include central cores with \(\Delta\) as high as 2%, compared to 0.3–0.6% for transmission fibers. As the wavelength grows longer, the optical mode progressively spreads out of the core, having more power in the trench and ring.

The effective index is, thus, forced to change rapidly with wavelength, and both the dispersion and the slope can be designed to be negative. One consequence of building-in the large waveguide dispersion is a necessarily small \(A_\text{eff}\) in the range of 15–21 μm2.

Because of the high \(\Delta\), which results in a softer glass at processing temperatures, the core ovality of a DCF is generally more difficult to control than that of a Tx fiber, and the stress asymmetry resulting from core ovality is also higher, resulting in somewhat higher PMD values.

However, good process control and the use of spinning allows modern DCFs to have PMD values of 0.1 ps/rt(km) or lower. The large Ge doping in the core increases Rayleigh scattering and the large change in material properties at core–trench interface elevates the loss of DCF to a typical range of approximately 0.5 dB/km.

Because of the higher loss and PMD, as well as the desire to minimize the size of the fiber bobbin, it is desirable to maximize the magnitude of the negative dispersion as practical of the fiber to minimize the length of DCM.

As a consequence, the dispersion divided by the fiber loss is often used as a figure of merit for DCM. Well-designed DCF should have a figure of merit larger than 200.


5.3. Full-Band Dispersion Compensation

To guarantee that a DCM precisely compensates the dispersion across the entire C- or L-band, it is necessary for the relative dispersion slope (RDS) of the Tx fiber and the DCF to be equal. Let us approximate the dispersion of an optical fiber as a linear function over a wavelength band,


where the center of the band is \(\lambda_\text{c}\) and \(D'=dD/d\lambda\) is the dispersion slope.



The RDS should be considered a fundamental parameter and figure of merit for a telecommunication optical fiber.

With reference to Fig. 5.7, consider that the transmission fiber is characterized by length \(L_1\) and dispersion \(D_1(\lambda)\), while the DCF is characterized by length \(L_2\) and dispersion \(D_2(\lambda)\). Then the condition for 100% dispersion compensation across the band of interest is that


must be satisfied for every value of \(\lambda\). Substituting Eq. (5.8) into Eq. (5.10) and regrouping terms yields the following expression:


In order for this expression to be true for any \(\lambda\) within the band, the two expressions in brackets must both be identically zero. Equating the two expressions in brackets yields

\[\tag{5.12}\frac{D_1'}{D_1}=\frac{D_2'}{D_2}\text{ or, using different notation, }RDS_1=RDS_2\]

This is the first requirement for precise wideband dispersion compensation. Matching the RDS of the Tx fiber and the DCF to achieve wideband compensation is often referred to in the literature as ‘‘slope compensation’’ or ‘‘slope matching.’’ DCMs, which offer 100% slope-matching across a band, are then referred to as ‘‘dispersion slope compensating modules’’ (DSCMs).


Figure 5.7  Typical compensated span in an amplified system for a long-haul DWDM link with erbium-doped fiber amplifier (EDFA) technology, in which the dispersion compensation fiber (DCF) in module is placed in a two-stage EDFA.


5.4 Requirement for Low Residual Dispersion

The second requirement for precise wideband dispersion compensation is that the Tx and DCFs have a low RDS. The assumption in Eq. (5.9) that the fiber dispersion curve \(D(\lambda)\) is linear is more valid for the Tx fiber than for the DCF.

In reality, \(D_2(\lambda)\) may have significant quadratic and cubic terms in the Taylor expansion. For 40-Gbps transmission line design, it is useful to define the relative dispersion curvature \(RDC=D^"/D\), which can also be matched between Tx fiber and DCF.

The term residual dispersion refers to the net dispersion of a compensated span or a concatenated series of compensated spans as illustrated in Fig. 5.7. If the RDS values of the Tx fiber and DCF have been matched, then the residual dispersion will be essentially parabolic across a band.

In cases in which the RDC can also been matched, then the residual dispersion takes on a cubic shape. It is this residual dispersion over the target transmission distance that must be compared to the dispersion tolerance of the receiver.

It can be shown that the available bandwidth (over which low residual dispersions are achieved) for a Tx fiber and a slope-matched DCF varies inversely with RDS. Essentially, DCFs with lower residual dispersions tend to have lower curvature \(D^"(\lambda)\). Empirically, the optical properties of DCF with low RDS values also tend to be more robust to manufacturing variation.

Thus, the RDS of the Tx fiber becomes a critical parameter in determining how well the dispersion of the link can be compensated. From Eq. (5.9), it is obvious that higher dispersion and lower slope facilitate the design of a matching dispersion compensation solution.

As a rule of thumb, it is relatively easier to design and manufacture DCMs with tight residual dispersion tolerances for Tx fibers with RDS values well below 0.01 nm-1, while residual dispersions will tend to be high and PMD and insertion loss penalties greater for fibers with RDS values significantly above 0.01 nm-1.

Referring to Fig. 5.1, fibers with high dispersion (~17-20 ps/nm-km) near 1550 nm, as well as those with medium dispersion (~7-8 ps/nm-km) and lower dispersion slopes, will fall into the former category.

A fiber with a lower value of dispersion (~4-5 ps/nm-km) at 1550 nm will fall into the latter category if its slope is high. A lower dispersion Tx fiber with low slope can still have good dispersion slope matching with a DCF, while maintaining lower module loss and PMD, if the RDS is not higher than 0.01 nm-1.


5.5  Factors Affecting Nonlinearity

Nonlinearity can arise in optical fiber transmission from the cabled Tx fiber itself, the amplifier, or in the DCM. The fundamentals of fiber nonlinearity and the manifestations in communication links are well summarized in the literature.

For a Tx fiber, the important parameters affecting nonlinearity are the nonlinear index \(n_2\), the effective area \(A_\text{eff}\), the zero dispersion wavelength \(\lambda_0\), and the level of local dispersion.

The value of \(n_2\) varies linearly with the overlap integral of optical power and the Ge-doping profile in the core of the waveguide. It is intuitively obvious that nonlinear interactions scale inversely with \(A_\text{eff}\). The nonlinear coefficient \(\gamma\propto{n_2}/A_\text{eff}\) combines these two parameters in a useful ratio.

It has been well understood since the early 1990s that DWDM transmission requires nonzero dispersion along the signal path to suppress FWM between channels, thus requiring that \(\lambda_0\) be well away from the transmission band.

Moving away from \(\lambda_0\), FWM products between channels are suppressed approximately as \(1/D^2\), requiring about 2 ps/nm-km across the signal bands in case of 10 Gbps. This consideration guided the development of the G.655 NZDFs in Fig. 5.3.

XPM between channels is left as the dominant source of nonlinearity for 10-Gbps transmission. The local dispersion of the Tx fiber does play a role in suppressing XPM in conjunction with a properly designed dispersion map.

In dispersion management, the optimum dispersion map prescribes the levels of pre-, post-, and in-line optical dispersion compensation so that the pulse collisions that generate non-linear interactions are minimized, on average, near the beginning of each span where optical power is maximum and within the DCM.

Similarly, the level of local dispersion in the Tx fiber also affects the magnitude of penalties due to SPM, XPM, and FWM between pulses within the same channel that are the dominant source of nonlinear penalties at 40 Gbps.

A somewhat higher local dispersion offered by the G.656 fibers can be useful, with the proper dispersion map, to further suppress these effects in the more demanding applications.

It can be shown from a simple analysis of dispersion curves shown in Fig. 5.3 that \(\lambda_0\) is inversely proportional to the fiber’s RDS. Thus, a lower RDS value will correlate with a greater bandwidth over which nonlinearity is suppressed.

Another contribution to nonlinearity arises from the DCM, which contains kilometers of fiber with larger values of \(\gamma\) (on the order of 5W-1km-1) compared to Tx fiber (on the order of 1.5W-1km-1).

The length of DCF necessary to compensate the Tx fiber scales with the length of the span (typically 60–100 km) and the dispersion of the fiber in the operating band. Full-slope compensation of fibers with a high RDS generally also requires longer lengths of DCF in the DCM.

Significantly, G.652 fiber has far more dispersion in the C- and L-bands than is necessary or useful for suppression FWM or other nonlinear effects, leading to excessive loss, PMD, and nonlinearity in the DCM.

For example, in the case of full C-band compensation of an 80-km span of SSMF (17 ps/nm-km at 1550 nm), approximately 35% of the nonlinear phase shift will occur in the DCM.

For the case of a low-dispersion (4.5 ps/nm-km at 1550 nm), low-slope (0.045 ps/nm2-km at 1550 nm) fiber such as TrueWave RS, less than 10% of the nonlinear phase shift for the 80-km span occurs in the DCM for full C-band compensation. A comparison of the impact of fiber dispersion on module loss and PMD is made in Table 5.2.


5.6 Impairments Affecting Raman Amplification

The advent of Raman amplification has drawn attention to the importance of FWM effects between the high-power Raman pumps and between the DWDM signals and high-power Raman pumps.

Essentially, it is desirable to place \(\lambda_0\) at a shorter wavelength than any signal or pump wavelength used in the system. For example, studies of Raman amplification led to the observation of FWM products between counter-propagating high-power pumps at 1429, 1447, and 1465 nm and C- and L-band signals when \(\lambda_0\) of the Tx fiber is very close to 1500 nm.

Significant noise peaks due to FWM were observed near 1528, 1548, and 1568 nm, causing OSNR degradation. These deleterious features did not appear for a Tx fiber with \(\lambda_0\) < 1405nm to the short wavelength side of the Raman pumps. Again, a low RDS value correlates with a greater bandwidth over which FWM impairments are suppressed.


5.7 Systems Implications of Tx Fiber PMD

In system work, one needs to account for both the PMD of each component of the transmission line shown in Fig. 5.1. A general rule of thumb is to allow half the system PMD for the fiber and half for the components.

Although the components may have a relatively fixed DGD, components in the middle of a system span appear to have a statistically varying DGD because of the fiber that surrounds them.

The total pulse spreading resulting from PMD is usually allocated no more than one-tenth of the bit period. Hence, for a 40-Gbps transmission system, the total pulse spread due to PMD could be as high as 2.5 ps. For a system of 1000 km length, the fiber PMD coefficient could be no higher than 0.04 ps/km1/2.

A more sophisticated analysis estimates the power penalty due to DGD for a given modulation format at a fixed line rate (power penalty is the increase in signal power required to overcome the impairment). For example, an NRZ system roughly suffers an eye-closure penalty of

\[\tag{5.13}\text{Penalty (dB)}=15[\text{DGD}*\text{bit rate}]^2\]

so for a 1-dB penalty, DGDmax = 0.26/(bit rate). This DGD = DGDmax value is taken at the limit at which the system fails. Because of the statistical nature of DGD, the actual instantaneous value of DGD could be far above the PMD value for the same system (since PMD is the average of the DGD).

In fact the maxwellian tail of the DGD distribution extends to infinity (though with very small probability density). The system designer uses the DGDmax to find the probability of instantaneous DGD causing the system to fail.

Knowing an acceptable outage probability, the overall PMD is adjusted downward to guarantee that DGDmax is far enough out on the maxwellian tail so that system failure probability is kept within the design limits.

It can be shown that the outage probability of 10-6 (or 30 seconds per year) requires a PMD value that is less than one-third of DGDmax. Some examples of PMD fiber requirements based on this analysis are shown in Table 5.1.


Table 5.1  Optical fiber PMD requirements for future telecommunications needs


5.8 Summary of Design Principles

Based on the previous discussion, a Tx fiber with reduced dispersion (relative to G.652 fiber) and low RDS offers system advantages in wideband dispersion compensation, reduced attenuation, PMD, and nonlinear penalties for the DCM module, avoidance of nonlinear impairments in distributed Raman amplification, as well as high efficiency for Raman gain.

The Tx fiber PMD must also be controlled to ultralow levels per Table 5.1 to take advantage of a fiber designed to these principles.


6. Design of Nonzero Dispersion Fibers

The design process for NZDFs can be summarized by the following simplified steps:

  1. Determine an acceptable compromise between the relevant fiber transmission properties that will support the intended applications. This primarily requires determining a mutually attainable set of values for the MFD, dispersion and dispersion slope within the transmission band, and macrobending and microbending loss sensitivity. This is normally accomplished using waveguide modeling tools.
  2. Determine the radial shape of the refractive index profile that yields the desired transmission properties.
  3. Ensure that the index profile can be robustly manufactured with the available preform fabrication process.

Because the physics of optical fiber transmission and the state of the art of glass processing technology will not always allow the simultaneous achievement of the targets and constraints, the process by nature is a compromise.

It would be impractical to implement this process without simulation tools that enable a ‘‘virtual’’ search of the fiber design parameter space for acceptable solutions.

Typically, finite element techniques are employed to solve the Helmholtz equation for a given waveguide structure and determine values for the pertinent fiber properties. Optimization techniques can be applied to the design search to locate solutions that best fit the design targets while maintaining the appropriate index profile constraints to ensure that the solutions obtained are realizable.

However, the uncertainty introduced by the imperfect nature of the practical physical models of optical fiber behavior will, in general, make the process an iterative process of modeling, prototyping, and performance testing, and subsequently repeating the process loop to ‘‘tune’’ a design to an acceptable optimum. In particular, it is challenging to make highly accurate absolute predictions of bending losses and cutoff from first principles. 


6.1 Fiber Transmission Parameter Tradeoffs 

A complicating factor of the fiber design problem is that there are several coupled problems that must be simultaneously solved when searching for an optimum set of fiber transmission parameters.

We would like to design fibers that minimize the impairments resulting from nonlinear propagation effects, but we may simultaneously desire that the fibers be robust to cabling and environmental changes while providing efficient low-noise distributed-Raman gain. This combination of desirable fiber properties puts conflicting demands on the size of the MFD.

Further, we would like flexibility to engineer the fiber dispersion curve (zero dispersion wavelength and dispersion slope) so that we

  1. Maintain dispersion within an optimum range across the entire transmission band to help mitigate some nonlinear impairments by braking phase matching conditions.
  2. We also want to have low RDS, to match the available dispersion compensating fiber technology so that linear dispersion can be optimally compensated over a broad bandwidth.

However, obtaining low-dispersion slope tends to preclude achieving simultaneously both large MFD and insensitivity to microbending loss.

In general, to minimize the transmission performance degradation caused by nonlinear propagation effects, large MFD and \(A_\text{eff}\) are desirable. However, the macrobending and microbending loss sensitivities have strong dependencies on MFD and tend to favor small MFD.

Raman gain efficiency is inversely proportional to \(A_\text{eff}\) and the square of MFD, so small MFD will increase Raman gain efficiency and lower the Raman pump power required for a desired value of gain. The dependency of nonlinear impairments, Raman gain efficiency, and bending-induced loss sensitivity on MFD requires finding the appropriate balance of these properties that support the performance requirements of the intended application.

The chromatic dispersion of a silica-based fiber is a function of the dispersion of the silica and the dispersive effects associated with propagation within the waveguide (i.e., waveguide dispersion).

In the wavelength region where a fiber is single moded, the waveguide dispersion of the LP01 mode is normally negative and the material dispersion is normally positive.

The waveguide dispersion effects result from the mode becoming less confined to the high index of refraction core region and spreading out into the lower index of refraction, faster phase velocity, cladding material as wavelength increases.

As wavelength increases, the LP01 mode \(\beta\) decreases, the group velocity increases and waveguide dispersion becomes more negative. Engineering the shape of the dispersion curve is primarily a process of designing a multilayer index profile shape so that the LP01 mode transitions from regions of high-index, slow-phase velocity to regions of lower index faster phase velocity so that the group velocity of the mode and, therefore, the dispersion are controlled.

Superimposed on the dispersion curve engineering is the need to maintain the MFD at the signal wavelength, the bending sensitivity and single-mode operation at desired values.


6.2 Realizability, Manufacturability, and Scalability 

When engineering a particular waveguide design, one needs to consider that the index profile shape must be realizable using the available manufacturing processing.

Typical considerations are the availability of dopants and related glass processing to achieve the required index delta of the various waveguide regions, the ability of the process to fabricate the required shape of the index layers, the ability of the process to achieve the required tolerances on the index levels and shapes, as well as the tolerances on radial dimensions.

Most modern commercialized transmission fibers are doped with germanium to raise the refractive index, to relative deltas \(\Delta\) up to about \(1\%\); while fluorine doping is used to depress the refractive index, to \(\Delta\) down to about \(-0.3\%\).

The realizability of a design will depend on the fabrication process available for manufacturing. The VAD process is well suited for fabricating the step-index core shapes typically used for SSMF designs.

However, the VAD method has been used to fabricate the more intricate waveguide shapes required for more advanced designs such as G.653 and G.655 fibers.

The first-generation VAD G.653 fibers were made with relatively simple waveguide shapes, as shown in Fig. 5.8.


Figure 5.8. Simple core-pedestal index profile shape used for moderate \(A_\text{eff}\) and dispersion slope nonzero dispersion-shifted fiber (NZDF) often made with the VAD process.


The core torch is carefully controlled to fabricate a relatively high index, alpha profile shape central core region, with alpha typically less than about 10.

A cladding torch with GeCl4 vapor is used to form the raised index ring region adjacent to the core. This relatively simple waveguide shape has been used for G.653 and G.655 fibers with moderate \(A_\text{eff}\) (~50-55 μm2) and moderate dispersion slope (~0.07-0.08 ps/nm2-km).

More complicated radial waveguide shapes are generally required to allow further enhancement in the \(A_\text{eff}\) or further tailoring of the dispersion curve. Figure 5.9 shows an index profile shape that provides an increase in the effective area while maintaining the bend sensitivity near that of SSMF.


Figure 5.9.  Core-ring index profile shape used for large \(A_\text{eff}\) nonzero dispersion-shifted fiber (NZDF).


The dispersion slopes associated with this family of designs can vary from moderate to high, ranging from 0.052 to 0.09 ps/nm2-km. The three-layer core structure includes a high-index, low-alpha, central core, surrounded by a narrow annular region with index that is nearly matched to the cladding index.

This structure is then surrounded by a second annular region with index raised above the cladding index. Fabrication techniques that build the index structure by radially layering multiple deposition steps, such as the OVD, MCVD, or PCVD methods, have been successfully used for fibers of these designs.

It is also possible to use the VAD method when multiple deposition torches (as many as five torches have been used with some success) are employed.

Figure 5.10 shows the index profile shape generally used in tailoring the dispersion curve to provide low dispersion slope.


Figure 5.10.  Three-layer index profile with core, trench, and ring layers used for low dispersion slope nonzero dispersion-shifted fiber (NZDF) with moderate \(A_\text{eff}\).


A large contrast between the phase velocities of the material in the central region and the first annular region is required. To enhance this contrast, the index of the first annular region is reduced below that of the cladding by doping with fluorine.

In general, fibers with low dispersion slope (<0.05 ps/nm2-km) must have moderate values for \(A_\text{eff}\) (<60 μm2) in order to have bending loss properties that allow placement in high fiber count, low size cable structures.

The MCVD and PCVD processes are more amenable to low-cost fabrication of depressed index layers than are the OVD and VAD processes. However, multiple VAD or OVD deposition, dehydration, doping, and sintering steps may be employed to fabricate annular fluorine-doped regions embedded within germanium or pure silica regions.


6.3 Low-Dispersion NZDFs

The family of G.655 fibers with low dispersion was co-developed as research progressed on 10-Gbps DWDM systems in the mid-1990s.

This class of fiber attempts to reduce the excess dispersion of G.652 SSMF in the C- and L-bands while maintaining sufficient dispersion at the lower end of the C-band to fully suppress FWM.

Removing the excess dispersion reduces the length of fiber in the DCM and thus reduces DCM loss, PMD, and nonlinear penalties. Two widely deployed commercial fibers make up this category, as shown in Fig. 5.11.


Figure 5.11. These two fiber designs manage the tradeoffs noted in Fig. 5.2 differently. One maximizes \(A_\text{eff}\), and the other minimizes the dispersion slope. The larger \(A_\text{eff}\) fiber offers the advantage of launching approximately 1.0–1.5 dB more power into the span at a fixed level of Tx span nonlinearity, in cases in which additional optical signal-to-noise ration (OSNR) is required. The tradeoff is that a higher slope fiber has a larger relative dispersion slope (RDS) and is generally less well suited for broadband compensation, incurring greater penalties and more residual dispersion for a slope-matched dispersion compensation module (DCM).


They have approximately one-quarter of the dispersion of G.652 SSMF, so both necessarily have lower values of \(A_\text{eff}\) as well. One fiber has a larger \(A_\text{eff}\) (~72 μm2) but higher dispersion slope (~0.09 ps/nm2-km) because of the tradeoffs discussed previously.

The other fiber has a moderate \(A_\text{eff}\) (~54 μm2) but a low slope of 0.045 ps/nm2-km and maintains cabled cutoff less than 1260 nm. The lower slope fiber requires the F-doped trench profile design, as shown in Fig. 5.10.

Both fibers are available with a standard LWP specification at 1383 nm. In fact, 16-channel x 10-Gbps CWDM transmission has been demonstrated over the low-slope G.655 fiber.

These fibers offer a direct benefit over G.652 SSMF in the area of uncompensated metro transmission. Because the dispersion at 1550 nm is four times lower than at SSMF, it is possible to transmit four times the distance without compensation; distances of 320 km have been demonstrated at 10 Gbps by systems vendors and more than 400 km at 2.5 Gbps.

This not only offers the cost benefit of eliminating the DCM but also enables the use of simpler cheaper EDFAs. The added loss of a good metro DCM, perhaps 3.5 dB for an 80-km span of G.652 SSMF, is usually inserted at a midstage access point in a two-stage EDFA, as shown in Fig. 5.12.

Thus, by removing the DCM, a simpler, lower cost amplifier can be employed as well. In this way, the total system cost can be reduced by 8% for a 2000 km, 10G network using low-dispersion G.655 fiber.


Figure 5.12.  A typical amplified and compensated span includes a dispersion compensation module (DCM) at the midstage access point of an erbium-doped fiber amplifier (EDFA). Use of a low-dispersion G.655 fiber allows elimination of the DCM and simplification of the EDFA by eliminating the second stage for transmission up to 320 km.


Laboratory results with optical duo-binary modulation and a maximum likelihood sequence estimation (MLSE) equalizer at the receiver indicate that low-dispersion NZDFs can support 10-Gbps transmission over 1200 km without in-line compensation using simplified EDFAs.

In one demonstration, DCF at the transmitter and duo-binary modulation were used in conjunction with an MLSE receiver to transmit 50-GHz spaced channels over 1200 km of the high-slope G.655 fiber in Fig. 5.11.

In a different approach, chirp-managed lasers with duo-binary modulation were used in conjunction with an MLSE receiver to transmit 50-GHz spaced channels over 1200 km of TrueWave RS, which is the low-slope NZDF in Fig. 5.11.

The low-dispersion fibers of Fig. 5.11 have been extensively deployed in regional and long-haul networks around the world starting in the late 1990s. Transmission experiments using TrueWave RS have demonstrated that the lower dispersion across the C-band remains sufficient to suppress nonlinearities for high-bit-rate DWDM applications.

For example, 40- by 40-Gbps channels were transmitted over 5- by 100-km spans of TrueWave RS in the C-band using a slope-matched DCM and hybrid Raman/EDFA amplification.

An early demonstration of a commercialized all-Raman transmission system demonstrated 64- by 40-Gbps channels over 1600 km of TrueWave RS between 1554 and 1608 nm, with 50- by 40-Gbps channels over a slightly shorter distance for the high-slope G.655 fiber.

This system avoids the lower C-band to circumvent difficulties associated with (1) compensating the lower C-band for fibers with high RDS and (2) applying Raman amplification in fibers with \(\lambda_0\) too close to the signals.

The beneficial impact on the DCM properties from reducing Tx fiber dispersion is shown in Table 5.2.

To illustrate the principles, values are drawn from a family of commercial slope-matched DCMs intended for precise compensation applications. Comparing the 17-ps/nm-km value for G.652 Tx fiber in the first line of Table 5.2 with the 4.5-ps/nm-km value for G.655 fiber in the second line shows the impact of reducing the length of DCF in the module on both loss and PMD.

The Tx fiber with RDS ~ 0.01nm-1 does suffer a modest penalty in residual dispersion. The Tx fiber with RDS ~ 0.02nm-1 suffers a larger penalty in residual dispersion and loses the benefit of lower dispersion with regard to reducing PMD in the DCF.


Table 5.2. Impact of Tx fiber properties on DCM properties


6.4 Medium-Dispersion NZDFs

The fibers that most closely meet the design principles for enabling precise dispersion compensation, reducing total system nonlinearity, and promoting efficient use of Raman amplification are the ‘‘medium-dispersion’’ NZDFs (MDFs) that meet both the G.656 and the G.655 ITU specification.

Figure 5.13 shows dispersion curves for two commercially deployed examples. To reduce the RDS and maintain a low \(\lambda_0\) while balancing the dispersion slope with effective area as required by Fig. 5.2, the 1550-nm dispersion should be in the range of 7–8 ps/nm-km.


Figure 5.13.  Dispersion curves for two G.656 ‘‘medium dispersion fibers.’’ The fiber shown by the solid line has a 1550-nm dispersion, relative dispersion slope (RDS), and \(A_\text{eff}\) of 7.3 ps/nm-km, 0.0055nm-1, and 55 μm2, respectively. The fiber shown by the dashed line has values 8 ps/nm-km, 0.0065 nm-1, and 63 μm2, respectively.


By reducing the dispersion slope in the C-band to 0.04 ps/nm-km, the TrueWave REACH LWP fiber shown in the solid line achieves a nominal \(\lambda_0\) ~ 1385 nm.

This represents a significant reduction of the excess dispersion of G.652 SSMF while making the entire S-, C-, and L-bands available for signals and Raman pumps, by keeping \(\lambda_0\) < 1385 nm.

The combination of medium dispersion and low RDS gives this class of fibers a moderate \(A_\text{eff}\) that enables efficient, cost-effective application of Raman gain. Numerous record-setting experiments for high-capacity, long-distance product experiments used transmission lines based on MDFs.

From Table 5.2, it is clear that reducing the excess dispersion of the G.652 SSMF while maintaining a low RDS allows the corresponding DCM to have better wideband residual dispersion, lower insertion loss, and lower PMD.

The nonlinear performance of the DCM is also improved. It has been shown that the low RDS ~ 0:0055nm-1 of the TrueWave REACH fiber supports the design of an industrialized, full C+L band slope-matched DCM with \(\pm0.15\) ps/nm-km residual dispersion.

This device was the basis for a demonstration of 80 channels of 40-Gbps transmission over the C- and L-bands of TrueWave REACH using a single wideband DCM plus all Raman amplification.

To enable wideband and high-bit-rate systems, it is important for the fiber design to aid the system designer in getting the maximum benefit from Raman amplification.

The combination of low RDS and medium dispersion leads to a moderate \(A_\text{eff}\) and enhanced Raman gain efficiency. For this higher Raman gain to be used effectively in system design, the Raman gain coefficient must have a tight manufacturing distribution.

This, in turn, requires good manufacturing control over \(A_\text{eff}\) as well as water peak loss in the Raman pump region. If the water peak loss is reduced from 0.5 to 0.35 dB/km at 1383 nm, then the attenuation at 1410 nm is correspondingly reduced from about 0.335 to 0.285 dB/km.

Figure 5.14 shows both the measured manufacturing distribution of the Raman gain efficiency CR of 16,000 km of TrueWave REACH LWP fiber, as well as Raman gain curves for TrueWave REACH compared to G.652 SSMF fiber. The tight distribution of Raman gain around the mean supports practical systems engineering requirements.


Figure 5.14.  Manufacturing distribution of Raman gain for 16,000 km of TrueWave REACH (left) and Raman gain curves for G.656 TrueWave REACH versus G.652 SSMF (right). TrueWave REACH has an \(A_\text{eff}\) = 55 μm2, compared to 85 μm2 for SSMF.


The higher peak Raman gain efficiency of the G.656 fibers compared to SSMF makes it possible to use less pump power to achieve a targeted Raman gain. Although Raman amplifiers have proven reliability, prudence is still recommended with high-power transmission.

The use of lower pump power can improve thermal management in equipment design and avoid power levels that raise the possibility of connector end-face damage or burning of the coating at incidental fiber bends.

Alternatively, a higher Raman efficiency allows significantly higher Raman gain for a fixed pump power level. This situation arises where the 500-mW IEC Class IIIb laser safety limit is of concern.

Consider the situation where two copropagating pumps and three counter-propagating pumps are used, but the total power including signals is constrained to 500 mW at each end of the span.

A span of the G.656 fiber with \(A_\text{eff}\) = 55 μm2 has 50% higher peak Raman gain efficiency and supports more than 16 dB of flat Raman gain, whereas the SSMF supports just less than 12 dB with similar gain flatness.

For a typical 100 km span with 21 dB total attenuation, the optimum balance of EDFA amplified spontaneous emission noise and Raman multipath interference noise performance occurs with about 16 dB of Raman gain.

The ability to employ high levels of Raman gain with cost and power efficiency without noise from nonlinear impairments makes possible very high capacity transmission in the traditional C- and L-bands, as well as novel architectures for lower cost systems in demanding applications.

In the area of high capacity-distance product transmission, 160- x 42.7-Gbps channels spaced at 50 GHz have been transmitted over 3200 km of TrueWave REACH using the single C+L DCM, for a spectral efficiency of 0.8 bps/Hz and capacity-distance product of 20 Pbps*km.

It has been shown experimentally that the moderate \(A_\text{eff}\) of the NZDF Tx fibers, responsible for the larger Raman gain, presents only a small XPM penalty across a wide range of launch powers when mixing 10 and 40G channels in a system with industrial margins.

It is also possible to use high levels of co-pumped Raman gain to accomplish a doubling of the span length from about 100 to 200 km, sometimes known as ‘‘hut skipping,’’ for ultra long haul 10G systems.

Adding large Raman co-gain allows one to reduce the launch power and thus the average signal power over the first ~ 40km of the span where the power is highest.

This reduces the impact of Kerr non-linearities and thus reduces the required OSNR to attain the targeted bit error ratio, such that 40 channels x 10.66 Gbps transmission over 12- x 200-km spans of TrueWave REACH fiber was accomplished with 14.5 dB of co-pumped gain, 22 dB of counter-pumped gain, plus EDFA.

A related challenge is the case of un-repeatered transmission over a very long, single span to a remote area with amplifiers and DCM only at the terminals. In this case, 20- x 10-Gbps channels were transmitted over 300 km (or 64 dB of loss) on TrueWave REACH fiber, using 25 dB of counter-gain, 12.5 dB of co-gain, plus EDFA.

A study on the TrueWave REACH fiber (chromatic dispersion curve shown as the G.656 #1 in Fig. 5.3) has demonstrated PMD performance that will enable transmission at the highest bit rates (40–80 Gbps) shown in Table 5.1.

This work involved measurements of PMD from the fiber manufacturing process, through the cabling process and into field installation. The statistical performance of links formed from fiber at each stage in the process was evaluated so that link design values of installed cabled fiber could be inferred from measurements on uncabled factory fiber samples.

This demonstration verifies that a commercially available G.656 fiber has low enough PMD to handle next-generation transmission networks.


7. A New Paradigm in Transmission Line Design

Instead of placing the DCF in a module, it is possible to design a compensation fiber that can be cabled. In the former case, loss and nonlinearity are added to the span by the DCM without transporting the signal any closer to the end terminal; in the latter case, the loss and nonlinearity associated with dispersion compensation also contribute to system reach.

Submarine systems have traditionally been built from spans in which both positive and negative dispersion fibers were cabled. A span comprising a positive and a negative dispersion fiber with equal RDS—both fibers being cabled—is known as a dispersion-managed fiber (DMF) span.

DMF spans were first deployed in a 10-Gbps trans-Pacific submarine fiber link, requiring extremely precise dispersion compensation over 96 channels.

The positive dispersion fiber was a super large area (SLA) fiber with dispersion +20 ps/nm-km and \(A_\text{eff}\) = 107 μm2, while the negative dispersion fiber was an inverse dispersion fiber (IDF) with dispersion approximately ~40 ps/nm-km and \(A_\text{eff}\) = 31 μm2, representative of the UltraWave fibers.

The accumulation of nonlinear penalties can also be minimized by employing DMF spans. The first two-thirds of each span is an SLA fiber, to accommodate a high launch power, while the greatly attenuated signal traverses the IDF fiber with smaller \(A_\text{eff}\) in the final third of the span.

Submarine experiments have also demonstrated Raman amplification with DMF spans. DMF spans represent an optimized transmission line for ultralong haul terrestrial routes as well as submarine systems.

In the past few years, exceedingly high capacity–distance product transmission experiments have been conducted with these fibers using both hybrid Raman-EDFA and all-Raman amplification.

In proposed terrestrial applications adapting the aforementioned submarine fibers, the first third of the span is SLA, the middle third is IDF, and the final third is again SLA.

This symmetrical arrangement allows launch from either direction as well as helping to reduce nonlinearity when employing Raman amplification. The highest power levels occur at launch in the first SLA segment (perhaps including co-pumped Raman gain), while the second highest powers occur in the final SLA segment due to counter-pumped Raman gain.

It can be shown that the exact position of the middle third of IDF fiber is not critical to amplifier noise figure improvement. DMF spans have also been demonstrated using the NZDF fiber labeled as G.656 #2 in Fig. 5.3 with a slope-matched negative dispersion fiber having ~16 ps/nm-km at 1550 nm.

In addition to enabling ultralong haul transmission over the widest possible bandwidth, DMF spans may also prove an enabling technology for all-optical networking.

The dispersion of a properly designed DMF span is inherently compensated across the entire C- and L-bands, providing complete flexibility in wavelength routing.


The next tutorial discusses in detail about electric-dipole transitions in real atoms


Share this post



Sold Out