Infinite bandwidth is a term often used to describe the capacity of fiber-optic networks. It’s an exciting concept even though it doesn’t exist in reality. Bandwidth allocation has long been used to differentiate service in fiber-optic networks. Starting from an understanding of a network’s finite traffic capacity and customer requirements, operators have created a tiered service structure with many levels of performance and corresponding tariffs.
The amount of capacity and its rate of expansion certainly have changed as a result of widespread optical fiber installation. At least 3,000 miles of fiber are being installed worldwide per hour, and the data-handling capacity of that cable is doubling each year.
Until the early 1990s, the capacity of a single fiber in a commercial network was limited by the need for periodic electro-optical regeneration. Although modulated light quickly transported information over the fiber, at the receiving end of each segment, the weak optical signal had to be converted to an electrical one. The electrical signal then could be amplified, reconverted to light, and launched into the next segment. In 1987, the Erbium-doped fiber amplifier (EDFA) was invented, which allowed direct optical amplification and removed the bandwidth bottleneck and expense of electro-optical regeneration.
At least as important as the speed and cost benefits they provide, EDFAs also support practical dense wavelength division multiplex (DWDM) systems. Regenerators amplify only the signal to which they are tuned. Consequently, a separate amplifier was required for each wavelength in a wavelength-multiplexed system. In contrast, an EDFA amplifies a 30- to 40-nm band of wavelengths, so only one is needed to simultaneously boost the levels of 40 or more data channels.
Commercial networks today combine EDFAs and DWDM to carry 400 Gb/s over a single fiber. In laboratory experiments, an aggregate 1.6-Tb/s data rate has been demonstrated, compared with the theoretical maximum data bandwidth of about 50 THz between 1,200 and 1,600 nm. No one can precisely say how all that capacity will be used, but at the current speed of data-rate increase, the 50-THz ceiling will become a limiting factor within 10 years.
In the meantime, optical network refinements are proceeding on several fronts, including the fibers themselves, new types of amplifiers, and new transmission modes.
Optical Fibers
Fibers are drawn at high temperature from preforms comprising precisely doped layers of glass. The typical 8- to 10-µm dia core may have been doped with Ge to raise its index of refraction while the much larger 125-µm-dia cladding will have a lower index of refraction.
Light launched into the central core will be contained because it is constantly bent back toward the center by the change in index of refraction as it approaches the cladding. In addition, the small diameter of the core of a single-mode fiber ensures that only one mode of transmission exists: there is only one path the light will follow.
As pure and as low-loss as today’s fibers are, their characteristics are not constant. For example, the index of refraction of silica glass changes slightly as a function of wavelength.
The phenomenon related to this behavior is called chromatic dispersion. Its effect is to change the width of a pulse traveling through a long fiber. Because the speed of light in glass is proportional to the index of refraction, parts of pulses having different wavelengths will travel at different speeds.
Chromatic Dispersion
Chromatic dispersion is particularly important in DWDM systems because of the interference that can be generated among wavelengths. From the late 1980s to early 1990s, fibers were produced with zero dispersion corresponding to the 1,550-nm region of minimum loss. This was the optimum solution until DWDM systems became practical (Figure 1 see below).
Figure 1. Chromatic Dispersion vs. Wavelength for Three single-Mode Fibers (Source: Lucent Technologies)
In a fiber where the signal center wavelength is transmitted near the zero-dispersion wavelength, all wavelengths near the center wavelength travel at the same speed, so single-channel transmitted pulses will distort very little. Unfortunately, because DWDM transmits many channels over a long fiber, interference products are generated between channels. At the receiving end, their amplitudes can be comparable to those of valid signals. And because DWDM wavelengths are regularly spaced, the so-called four-wave mixing (FWM) interference products are exactly coincident with actual channels.
Adding dispersion to a fiber causes wavelengths to travel at different speeds, termed walk-off, which significantly reduces FWM. Chromatic dispersion is needed, but overall it has to be zero if the pulses are to maintain their timing relative to each other and their shapes. One current approach alternates positive and negative dispersion sections of fiber to achieve an overall zero sum while at the same time reducing FWM.
If fiber with large amounts of dispersion has been used, special dispersion compensating fibers (DCFs) are available with typically -100-ps/nm-km values. This situation could arise, for example, from transmitting 1,550-nm light through legacy dispersion-unshifted fiber with zero dispersion at 1,310 nm. At 1,550 nm, the dispersion is about +17 ps/nm-km compared with non-zero dispersion shifted (NDSF) or reduced-slope NDSF fibers having about ±4 ps/nm-km or less.
Nonlinear dispersion effects occur because the index of refraction also is affected by the power transmitted through an optical fiber. Self-phase modulation (SPM) describes the frequency shift or chirp caused to the pulse’s leading and trailing edges by opposite changes to the index of refraction. The leading edge slows down and the trailing edge speeds up corresponding to increased power at the beginning of the pulse and decreased power at the end.
As a result, a second-order effect called modulation instability (MI) can occur. If the amount of positive dispersion is great enough, the longer wavelengths of the slowed down leading edge can be overtaken by the relatively faster, shorter wavelengths of the trailing edge.
Clearly, there are at least two effects operating in opposition. Positive chromatic dispersion tends to stretch pulses. SPM-induced pulse asymmetry, if severe and if chromatic dispersion is sufficient, can cause MI, effectively damaging pulses.
By careful design of the fiber’s dispersion and SPM, the pulse can be made to contract and expand as it propagates and arrive at the end of the link without broadening. This is based on the fact that the chirp induced by dispersion in a dispersion unshifted fiber has a negative sign compared to the positive one induced by SPM.
Polarization Mode Dispersion
Polarization mode dispersion (PMD) refers to the difference in the propagation rate for the two polarization states of light in a single-mode fiber. In long links, energy is randomly exchanged between the two polarization states because of small changes in the fiber’s properties and bending and twisting. Any noncircularity in the fiber core makes the index in the X axis different from the one in the Y axis. This provides two different propagation constants and consequently birefringence. PMD adds to pulse broadening caused by positive chromatic dispersion.
As an example of a method used to control PMD, Lucent has developed a patented manufacturing process that introduces several twists per meter into fiber as it is drawn. The twists force high coupling between the two allowable polarization states, producing a value for PMD of about 0.5 ps/km and minimizing its sensitivity to external perturbations.
Raman Amplification
Nonlinear effects need not all be bad. At sufficiently high power levels, light energy interacts with the actual structure of the glass. The result is absorption of energy and subsequent release at a frequency 13 THz lower—at wavelengths from 60 to 100 nm longer than the original incident light within the 1,200- to 1,600-nm window. This effect is Raman scattering and forms the basis of Raman amplification.
EDFAs require a specially doped fiber and specific pump wavelengths and only amplify a small part of the available spectrum. Raman amplifiers offer the flexibility of handling a relatively broad band of frequencies in the overall 1,200- to 1,600-nm window, provided only that the pump wavelength is about 100 nm shorter than the center of the band to be amplified. The pump signal is launched into the fiber at the receiving end and provides distributed preamplification as it propagates toward the transmitter.
Attenuation
In addition to refining the engineering of fiber dispersion, fiber purity also has improved. Only a few years ago, the usable wavelengths in a typical fiber were divided into bands including the conventional (C) band from 1,530 nm to 1,562 nm and the long wavelength (L) band from 1,570 nm to 1,610 nm. The C band corresponds to the wavelengths first addressed by EDFAs.
Transmission was not possible between the original dispersion-unshifted fiber zero-dispersion point at about 1,310 nm and the C band because of a large hydroxyl-related attenuation peak at 1,383 to 1,385 nm. New manufacturing techniques have reduced this peak so a broad window exists between 1,200 and 1,600 nm. This means that many more DWDM channels can be accommodated by a single fiber, considering that cost-effective transmitters and amplifiers are made available in those new wavelength ranges.
With the capability to design fibers for certain applications has come diversification. For example, nonlinear effects can be minimized by increasing the effective core area. Fibers have been developed that retain the important single-mode propagation characteristics but distribute power more effectively within the core. These fibers offer up to 25% longer reach between amplifiers because they support higher transmitted power. Alternatively, because a larger effective area reduces power density within the core, second-order effects can be minimized for the same power level and reach.
Non-zero dispersion shifted cables with very low negative dispersion are ideal for submarine DWDM applications. For example, Alcatel provides up to 40 channels at either 2.5 Gb/s or 10 Gb/s on each of 12 or 24 fiber pairs. The maximum capacity of this type of system is 9.6 Tb/s.
Alternatively, if only a single channel is involved per fiber, it can operate at the fiber’s zero-dispersion frequency, and much longer spans can be achieved. A 2.5 Gb/s data rate provides more margin for dispersion-related pulse distortion so longer spans are possible.
Testing
The complex behavior of optical fibers may affect test results, particularly for very long-distance runs having many amplifiers between the send and receive ports. In such an example, a fiber segment will be spliced or otherwise connected to one of the amplifiers, with the amplifier connected to the next fiber segment and so on. Each of the interfaces will produce a reflected wave and account for part of the overall loss budget.
Also, there will be other types of optical and electro-optical devices encountered in the overall network: DWDM multiplexers and demultiplexers, dispersion compensation devices, optical add/drop multiplexers, and EDFAs or Raman amplifiers. Depending on the type of test, the effect of the network complexity may or may not be significant.
To the device manufacturer, all the second-order and cross-effect anomalies are important. The device is tested to a Telcordia, IEC, or ITU specification, and the guaranteed performance margin directly relates to measurement uncertainty.
With the advent of DWDM, test instruments must deal with time, parameter amplitude, and the influence of wavelength. As a result, for many device performance tests, an optical spectrum analyzer or a high-resolution wavelength meter is recommended.
Especially as data rates increase, the effects of dispersion become more important. There are specialized chromatic dispersion analyzers and PMD analyzers that can determine the characteristics of a fiber network. Optical time domain reflectometers (OTDRs) are used to verify continuity and uniformity of a fiber run, including any splices or connectors along the way. Reflections present because of the imperfect match created by connections or impairments can be measured by optical return loss (ORL) meters.
Of course, a cable installer is unlikely to delve into many of these aspects. Hopefully, the test equipment he uses will enable him to distinguish among the different types of cable likely to be present and measure their continuity and loss.
On the other hand, the commissioning engineer may have to look more deeply when presented with a connected network that doesn’t operate quite as expected. For example, he may need to use a bit error rate (BER) test set and a variable optical attenuator to help determine the source of marginal performance.
When a network is operating correctly, the objective is to keep it working well. Reference measurements are invaluable in finding the cause of network degradation: It used to work, so something changed that stopped it from working.
The EXFO FTB-5320 Multi-Wavelength Meter module from EXFO Electro-Optical Engineering, part of the FTB-300 Universal Test System, is useful in this type of network maintenance because it measures channel wavelength, spacing, and power levels. Repetitive measurements are facilitated by single-button operation that automatically returns center wavelength, peak power, and signal-to-noise ratio for each carrier (Figure 2).
References
- Tyrone, B., Lee, X., and Vigot, S., Testing Requirements for Dense WDM Optically Amplified Links, Application Note, EXFO Electro-Optical Engineering.
- Refi, J., “Optical Fibers for Optical Networking,” Bell Labs Technical Journal, Lucent Technologies, Jan.-March 1999.
- Dowdell, E., High Data Rate Networks, The Latest Fiber Technologies For Long-Haul, Corning, presented at AMTC 99.
- Guide to DWDM Technology and Testing, EXFO Electro-Optical Engineering, 2000.
Acknowledgement
We thank Dr. André Girard of EXFO Electro-Optical Engineering for providing material for this article.
Published by EE-Evaluation Engineering
All contents © 2001 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
March 2001