Amplifiers at 1590 nm double the dwdm bandwidth
Dan Yang afc Technologies Inc.
Driven by the rapid increase in Internet use and the introduction of applications like digital television and videoconferencing, the demand for a significant increase in bandwidth is growing at an unparalleled rate. Only two years ago, the 16-channel dense wavelength-division multiplexing (dwdm) system was commercially introduced. Today it is already insufficient to meet high-capacity requirements for the transfer of information. Increasingly, products with up to 80 channels are starting to appear on the market. As they do, it`s clear that a race has begun among systems manufacturers to provide ultra-high-capacity dwdm systems.
Essential to making any increase in capacity a reality is the optical fiber amplifier, the key element and the enabler of dwdm systems. The traditional foundation for dwdm systems is the erbium-doped fiber amplifier (edfa). But, like the 16-channel dwdm system, the edfa may no longer provide enough bandwidth to meet escalating requirements. A new amplifier technology promises to increase the amount of usable bandwidth and enable the development of a new generation of high-capacity dwdm equipment.
edfas limit dwdm capacity
Amplifiers compensate for signal loss that propagates in optical fiber. The transmitting fiber and other branching components such as add/drop filters and wavelength-selecting elements may contribute to this condition.
Although still a recent technology, the edfa has rapidly eliminated the need for electrical regenerators, thus simplifying the whole lightwave transmission network. However, the usable bandwidth of edfas is only about 30 nm (1530 to 1560 nm)--i.gif., 30% of the total bandwidth of a transmission fiber where attenuation is at the minimum (1500 to 1600 nm).
In a wavelength-division multiplexing (wdm) system, channels (wavelengths) are laid one against another. Due to the limited 30-nm bandwidth and the desire to simultaneously transport as many channels as possible, the distance between each channel is very small, typically 0.8 to 1.6 nm. As a consequence, interchannel crosstalk, which results from a transfer of power from one channel to another, becomes one of the most important issues in designing dwdm communications systems.
Crosstalk can be categorized into two types--linear and nonlinear. Linear crosstalk originates in the channel-selection device (i.gif., demultiplexers) and depends on factors like interchannel spacing and the demultiplexing element used to select the channel. The imperfect nature of the demultiplexers means each channel admits a small portion of power from neighboring channels, thereby interfering with detection and increases the bit-error rate (ber).
Four-wave mixing in the silica fiber is the main source of nonlinear interchannel crosstalk. When three wavelengths co-propagate inside the fiber, a new wavelength is generated. In the case of a wdm system with equally spaced channels, these new channels coincide with existing channels. As a result, not only do the existing channels lose their power, but the new ones frequently mask detection by the receiver of the original signals.
It`s not difficult to conclude that the amplitude of the crosstalk is directly proportional to dwdm channel spacing, meaning the latter cannot be infinitely decreased. As a result, the actual dwdm system`s capacity is inherently limited by the edfa bandwidth.
Another 35 nm
edfa bandwidth must be extended. What`s needed is an amplifier that dramatically increases bandwidth and allows the control--if not elimination--of crosstalk problems.
Recent studies have shown that a broader-bandwidth fiber amplifier is possible and that amplification wavelength can be extended beyond 1600 nm. Three types of amplifiers have been proposed--edfas based on other glass materials such as fluoride or tellurite, the Raman amplifier, and the standard silica erbium fiber-based dual-band fiber amplifier (dbfa).
It is well known that edfas based on new glass materials such as fluoride exhibit a flatter and wider gain profile. But they have two major problems. First, they are not compatible with transmission fiber, and thus cannot be fusion-spliced. Second, they are sensitive to humidity and show large gain variation with temperature change. As a result, reliability, packaging, and manufacturing issues must be addressed before they can be widely deployed.
Raman amplifiers use the stimulated Raman scattering that occurs in silica fiber when an intense pump beam propagates through it. Due to its nonlinearity, the energy-transfer efficiency is very low. As a huge amount of pump power is needed, laser safety is an issue. Also the Raman amplifier cannot compete with the compact size and flexible features of the standard edfa, as tens of kilometers of transmission fiber are necessary to obtain any relevant amplification.
Compared to these two approaches, the silica erbium fiber-based dbfa is more attractive because of its similarity the conventional edfa. Erbium fiber can emit light beyond 1570 nm under 980- or 1480-nm pumping--a phenomenon less known than 1550-nm amplification--due to its extremely low efficiency. Researchers at Bell Laboratories (Murray Hill, NJ) and Nippon Telegraph & Telephone (ntt--Tokyo) have constructed experimental dbfas based on two sub-band amplifiers with silica erbium fiber. They also have demonstrated terabit transmission experiments in the lab using dbfas, and results appear to be very encouraging. But due to the low pump efficiency, up to 500 mW of pump power have been sent to pump erbium fiber by combining a master oscillator-pump amplifier and/or high-power pump lasers so that performance similar to that of edfas can be obtained. The first commercial high-efficiency 1528- to 1610-nm dbfa was introduced only recently at ofc `98.
As mentioned, dbfas consist of two separate sub-band amplifiers: One is the conventional 1550-nm edfa (1530~1560 nm) and the other is the 1590-nm extended band fiber amplifier (ebfa), which has an operating wavelength from 1570 to 1605 nm. When the 1550-nm edfa and 1590-nm ebfa are multi plexed/demultiplexed in a parallel circuit (see Fig. 1), they offer a total of more than 75 nm of bandwidth.
Off-the-shelf components
The 1590-nm ebfa uses standard off-the-shelf components, making large-volume manufacturing cost-effective. It can be integrated into exactly the same module package widely used by edfa manufacturers. This compatibility allows 1590-nm ebfas to be plugged directly into existing or new wdm systems, either to work as an independent amplifier or to extend the bandwidth of existing edfa systems.
Compared with the conventional edfa, the 1590-nm ebfa features several attractive aspects:
Flat gain as an inline amplifier: The gain flattening of edfas is another hot topic in dwdm application, especially in an inline-amplifier situation. An edfa has a gain peak at 1532 nm and a dip around 1538 nm (see Fig. 2). When cascaded for long-haul point-to-point transmission, the accumulated gain is much higher at 1558 nm (see Fig. 3). A massive effort has been put into developing gain-flattening techniques. Among the most common is passive gain-flattening by filters, including fiber gratings.
In contrast, 1590-nm ebfas exhibit a very flat gain over more than 35 nm of bandwidth, with a better dynamic range to resist input-level variation. When in a cascade configuration, the accumulated gain profile can stay relatively flat, which facilitates the gain flattening process. This feature is exceptionally attractive, especially for long-haul terrestrial or submarine dwdm systems in which thousands of amplifiers might be needed in a single transmission link.
Slow saturation: The 1550-nm edfa achieves deep saturation once the input signal exceeds about -5 dBm. That means the output will remain almost constant, even though the input level is still increasing. In the dwdm situation, assuming each channel transmitter provides about 0 dBm, with 16 channels, the total power is already 12 dBm. As with 0 or 12 dBm, the edfa booster outputs the same power; the 12 dBm of power is wasted. The 1590-nm ebfa saturates differently (see Fig. 4), with output power increasing almost linearly, even at input powers above 0 dBm. This situation provides a great opportunity to use the high-power laser transmitters that are already commercially available.
High-gain and low-noise figure: The noise figure and small signal gain of 1590-nm ebfas also are comparable to edfas: 20 to 25 dB over 35 nm of bandwidth, with a typical noise figure below 5 dB (see Fig. 5). These numbers satisfy the pre-amplifier requirement. Thus, the 1590-nm ebfa is suitable for receiver applications.
Solution to crosstalk
A major concern in dwdm systems is channel cross talk. The nonlinear 4-wave mixing problem is even more severe in dispersion-shifted fibers (dsfs). The standard singlemode fiber was originally designed for transmitting 1.3-micron light. It has a large chromatic-dispersion coefficient at 1.55 microns, about 18 psec/km-nm. Dispersion induces pulse broadening. As signals are transmitted by digital bits, the broadened pulses not only see a power decrease, but when spread beyond their allocated bit slot, they interfere with each other. As a result, receiver sensitivity is affected.
To decrease the dispersion effect at 1.55-micron transmission, dsfs were designed so the fiber minimum-loss point coincides with the zero dispersion wavelength and edfa gain band. When dwdm systems are introduced to transport high-capacity channels over dsfs, the very small chromatic dispersion of the fiber causes phase matching among different wavelengths, meaning the 4-wave mixing threshold is easily reached. As a result, dsfs suffer more severe channel crosstalk than the standard 1.3-micron transmission fiber.
The 1590-nm ebfa solves this problem by shifting transmission wavelengths into the 1590-nm band. In that region, the dispersion increases slightly--3 ps/(km-nm) at 1590 nm, instead of 0.05 ps/(km-nm) at 1550 nm--but remains much smaller than dispersion in the standard fiber. This dispersion is small enough to support high data rates but large enough to efficiently decrease 4-wave mixing.
The advantages of the new 1590-nm ebfa are therefore obvious. Used together with an edfa, it enables dwdm bandwidth to more than double, dramatically increases capacity for fiber-optic transmission, and provides more room for channels, thus decreasing or eliminating crosstalk. Employing only standard silica fiber components and regular pump lasers, while sharing the same circuit and packaging of conventional edfas, the 1590-nm ebfa is fully compatible with dwdm equipment. That means there is no difficulty in implementation--alone or in combination with an edfa--for dwdm transmission.
In addition, upgrading an existing dwdm system that uses an edfa with the 1590-nm ebfa is cost-effective, since it can be plugged into the systems to complement edfa bandwidth instead of replacing existing edfas. Therefore, the 1590-nm ebfa represents a major step forward in meeting the ever-increasing demands of high-capacity fiber-optic transmission systems to carry voice and data traffic and support Internet activities. u
Dan Yang is chief scientist at afc Technologies Inc. (Hull, QC, Canada).