Advanced fiber amplifier designs improve WDM system performance

Advanced fiber amplifier designs improve WDM system performance

Engineering studies continue to refine the output power, gain, noise, and materials characteristics of optical-fiber amplifiers for use in high-capacity, wavelength-division multiplexed networks

Bruce Nyman, Donald R. Zimmerman, and John R. costelloe

JDS FITEL INC.

The erbium-doped fiber amplifier (edfa) is an enabling technology for wavelength-division multiplexing (WDM) transmission systems. The development of these optical-fiber amplifiers for use in WDM systems has resulted in more-stringent design requirements for

amplifiers and their components. For example, the development of 32-channel WDM systems is driving the need for optical amplifiers that can provide as much as 25 nm of usable bandwidth.

For use in WDM systems that call for capacities greater than 40 Gbits/sec, the required amplifier bandwidth is still undergoing widespread study. For an 80-Gbit/sec system based on 10-Gbit/sec channels, only 8 wavelengths are needed, whereas 32 wavelengths are needed for 2.5-Gbit/sec channels.

The amount of gain flattening required for erbium-doped fiber amplifiers in high-capacity WDM systems and the methods used to obtain the required amplifier performance ultimately depend on a detailed analysis of the actual system.

Initial applications of WDM in transmission systems used coarse WDM technology and combined both 1310- and 1550-nm signals to double the capacity of installed fibers. To address the recent need for higher transport capacity without installing new fiber cables, narrowband WDM systems have been developed that use multiple wavelengths in the 1550-nm region. Currently, 8- and 16-channel systems are commercially available, and 32- and 40-channel systems have been announced for availability later this year. These systems provide a graceful way to increase capacity on installed fiber routes at a lower cost than installing new fiber. Additional functionality is expected when WDM systems evolve to in clude wavelength add/drop capability for optical networking.

An enabling technology for the development of narrowband WDM systems is the edfa. In its simplest form, an edfa consists of a length of erbium-doped fiber, a suitable-wavelength pump light source, and a multiplexer to combine the signal and pump wavelengths (see Fig. 1).

In amplifier operation, the light from the pump source excites the erbium atoms to stimulate emission of additional photons of the same wavelength as an incoming signal, thus resulting in amplification of the incoming light. The edfa has usable gain over the spectral region from 1530 to 1565 nm. To be simultaneously amplified, all channels of a WDM system must fall within this band.

In a regenerated system, the WDM wavelengths are demultiplexed and then individually regenerated before multiplexing for propagation along the transmission line. Due to the higher output powers available and the simultaneous amplification of many wavelengths that edfas allow, substantial numbers of optical regenerators can be eliminated from a route design. The reduction in number of system elements lowers system cost and improves system reliability.

Compared to edfas used in single-channel systems, edfas designed for WDM systems have additional requirements, of which the most important is the gain shape required for the system bandwidth. For example, some early commercial systems use a channel spacing of 200 GHz (approximately 1.6 nm) to facilitate the design and manufacture of demultiplexers and transmitters. For an 8-channel system, this leads to a system bandwidth of 12 nm. For 32-channel systems operating at a channel spacing of 100 GHz, the system bandwidth needed is 25 nm.

To control the amount of crosstalk leaking through the demultiplexer from adjacent channels, the difference in signal amplitudes of the channels at the receiver should be limited to less than a few decibels, even after a number of edfas have been cascaded. This limit severely constrains the gain shape requirements of the edfa. Some gain shape can be compensated by using source pre-emphasis. However, the usefulness of this method is limited, especially as the number of cascaded edfas is increased and as channels are added and removed.

The gain shape of an amplifier is mainly determined by the erbium-doped fiber used but is also affected by the wavelength response of the components. Fused-fiber components, although typically inexpensive, often introduce rolloff at the edge of the usable band. For example, some fused-fiber 1480/1550-nm WDM couplers have a cosine-like wavelength re sponse that peaks at 1550 nm, with a 2-dB rolloff by 1530 nm.

Additionally, the polarization-dependent loss (PDL) of the fused-fiber component generally varies with wavelength and leads to time-varying power levels for channels located at the edge of the band. However, both problems can be minimized through the use of dichroic-filter-based components. The loss variation in a typical dichroic 1480/1550-nm WDM is less than 0.1 dB, and the PDL of the device is less than 0.04 dB over the entire edfa wavelength range (see Fig. 2).

The gain shape of the edfa is also dependent on the operating conditions of the amplifier. The key parameters include pump power, input wavelengths, and power levels. The pump power is usually monitored and controlled to ensure a specific amplifier output power. The total output power and the output power per channel must be controlled to prevent optical nonlinearities from distorting the transmission signals and to avoid overloading the receivers.

WDM systems are designed to accommodate a wide range of fiber losses between edfas, without changing input conditions. The edfa must maintain its performance over all allowable combinations of input power and channel occupancy. And to add to the design challenge, these conditions might change over time as channels are added and removed. Various methods are available, however, to overcome these challenges and obtain gain-flattened amplifiers.

Erbium fiber composition is critical to edfa gain flatness. Currently, there are two options for the choice of host glass: silica or fluoride. Some manufacturers use fluoride-based fibers, whereas others continue to use the traditional silica-based fibers. Both choices entail various design considerations. However, the host glass that will win the majority of the market has yet to be decided.

Erbium-doped fluoride fiber amplifiers are available from only a few organizations, including Alcatel, Nippon telegraph & Telephone, Le Verre Fluoré, and Furukawa. Although fluoride fiber amplifiers exhibit better inherent gain flatness than silica fiber amplifiers, there are concerns about their long-term reliability. Fluoride-based fiber is hydroscopic and cannot be fusion-spliced to silica-based fiber. However, it can be hermetically coated to permit mechanical splicing to silica-based fiber.

Once the drawbacks are overcome, the performance attributes of fluoride-fiber-based amplifiers become more appealing. For example, these amplifiers have been demonstrated to achieve a 1.5-dB gain flatness in the 1532- to 1560-nm window. In addition, their spectral response is less sensitive to operating conditions than that of silica-based amplifiers. However, on the downside, fluoride amplifiers can only be pumped at 1480 nm, leading to noise figures that are at least 1 dB higher than those of 980-nm pumped silica-based amplifiers. This problem has been resolved in some designs by using a 980-nm pumped silica-based amplifier as the first stage in a multistage amplifier.

Various gain-flattening techniques can be applied to silica-based erbium-doped fibers. These techniques include using different fiber compositions, adding gain-compensating components, or using both techniques in multistage designs. In addition, various methods can be used to control the output power level.

The fundamental gain-flattening technique for single-stage, silica-based amplifiers is to change the composition of the erbium-doped fiber. The gain shape is affected by the concentration of the aluminum co-dopant in the host glass. For a single-stage amplifier with only a 20-nm bandwidth, this approach can provide a better gain flatness than can a fluoride-based amplifier. Another technique involves adding a gain-compensating device at the output of the amplifier. This device can be an interference-filter-based unit, a fiber grating filter, or a piece of specially doped fiber.

Interference-filter-based devices provide the greatest flexibility for filter design. These devices can be designed to compensate for both the 1532-nm gain peak and the slope in the 1550-nm region. Additionally, interference filters can be incorporated into devices based on grin lens technology, such as isolators and 980/1550-nm or 1480/1550-nm WDMs (see Fig. 3). These hybrid devices improve performance by reducing post-amplifier loss and simplifying manufacturing. Long-period fiber gratings can also be used to implement a gain-compensating filter. These gratings work by coupling the filtered light into the higher-order modes of the fiber.

Another design approach uses a specially doped optical fiber, most commonly samarium-doped fiber. The loss slope of this fiber is the inverse of the loss slope of an erbium-doped fiber. By splicing the correct length of samarium-doped fiber, the gain slope is compensated. Yet, for the single-stage amplifier, the main drawback of all gain-compensating devices is a reduction in the output power.

Output power reduction can be eliminated through the use of dual-stage amplifier designs. In single-wavelength applications, it is well-known that dual-stage amplifiers provide lower noise figures and higher output powers. For WDM systems, the amplifier typically uses a low-output-power, 980-nm pumped, first stage to obtain a low noise figure. The second stage is then designed as a power amplifier. The gain-compensating element is located between the two stages. An alternative is to use erbium-doped fibers with different dopants, such as phosphorus or ytterbium. These fibers can have gain shapes that are complementary to an aluminum-doped first stage.

Another concern for silica-based edfas is that these gain-flattening designs depend on the operating conditions of the amplifier. These conditions might differ from the nominal case due to variations in span loss or the number of channels transmitted. They are addressed through active control of the amplifier. The two main strategies are automatic-gain-control loops using control of the pump power and gain locking using an out-of-band laser.

The automatic-gain-control loop uses taps and detectors at the output of the amplifier to measure the signal power in each channel. Channels are identified via low-frequency modulation of the high-speed data. Based on the measured channel power and the number of channels, the pump power is adjusted accordingly. The other approach, gain locking, keeps the amplifier operating point constant. This condition is achieved by adding an additional signal to the amplifier input. The level of this signal is controlled to keep the total power input to the amplifier constant.

A novel approach to controlling the out-of-band signal level is to use the amplifier as the gain medium in a laser cavity. The laser wavelength is operated outside the signal frequency range. Since the output power of the amplifier is constant, there is competition between the signal and lasing wavelengths. As the signal power increases, the lasing wavelength power decreases. One advantage to this approach is that it can easily be implemented with two fiber Bragg gratings.

Although it may be unclear which technology will dominate the market, it is clear that wavelength-flattening amplifiers are required and will remain critical to WDM system development. u

Bruce Nyman, Donald R. Zimmerman, and John R. Costelloe are senior engineers at jds fitel Inc., Lightwave Products Group, in Eatontown, NJ.