By MEGHAN FULLER
According to market researcher Communications Industry Researchers Inc. (CIR-Charlottesville, VA), the worldwide amplifier market will grow nearly 80% in the next five years, jumping from $600 million in 2001 to more than $1 billion by 2006. While erbium-doped fiber amplifiers (EDFAs) are still the most commonly deployed, the long-haul (LH) and ultra-long-haul (ULH) segments are now seeing hybrid EDFA/Raman amplification in very long spans or where systems are particularly noisy. Added intelligence and increased functionality are also key trends in the LH amplifier space, but perhaps the biggest trend is an emerging need for amplification in the metro.
Christine Mulrooney, senior analyst of optical components at CIR, contends that LH and ULH networks will continue to be EDFA-based in the foreseeable future. However, she adds, "almost all of the equipment manufacturers that I talked to are using an EDFA/Raman hybrid. They put a Raman amplifier in line with an EDFA to boost the signal and reduce the noise."
Despite the proliferation of EDFAs, Raman amplifiers will be increasingly important going forward, particularly as they become more affordable. "The key driver from our customers and from our customers' customers is to reduce the cost of the network from a capital-expenditure and operational-expenditure point of view," says Stefan Rochus, product-line manager for EDFAs, Raman, pump modules, and pump lasers at Agere Systems Inc. (Allentown, PA).
In the near future, integration of passive and active components will be key to lowering amplification costs, he says, and while Raman technology is particularly well suited for integration, EDFAs are not. "You have 980- and 1,480-nm pump lasers. The 980 is based on gallium arsenide substrate, and the 1,480 pump lasers are based on indium phosphide," he explains. "Then you have passives, which are based on thin-film materials-glass or other crystals-and you have the erbium-doped fiber, which is a fiber-based device. So you are working on essentially four or five different platforms."
According to Rochus, there are a few companies actually considering an all-Raman system. When systems are upgraded to 40 Gbits/sec, Raman will play an increasingly important role because it provides a low noise figure, but to displace EDFAs for LH and ULH applications is highly unlikely in the foreseeable future, asserts Mulrooney. "Everyone that I have talked to is, at least for the next five years, doing some sort of EDFA," she explains, adding that all manufacturers in the LH and ULH segments are looking for more functionality in the amplifiers.
Mark Verreault, senior brand manager of optical long-haul at Nortel Networks (Ottawa, Canada), cites another major trend in the LH and ULH segments: Today's optical systems are going further and further. They also tend to be more photonic and are designed to minimize the amount of regeneration necessary to transmit voice and data traffic. As a result, he says, network engineers must monitor amplifier sites more carefully.
"In the past, we've been accustomed to breaking down the signal into its electrical format every 300 to 400 km, and then we can at least identify over that link the quality of every one of our channels. We had the ability, through the overheads-via SONET or other forms of protocol-to assess that," he explains. "So one of the trends we have done a lot of work on is putting in optical spectrum analyzers that monitor power and provide an estimate of the optical signal-to-noise ratio."
David Ronald, director of product management for optical subsystems at Altamar Networks (Mountain View, CA), claims the ability to set the gain or variable gain control is another emerging trend. "As you are changing the overall loading on the system, you want to be able to play with the gain," he says, "and it's a nice thing if you can do that remotely."
Several manufacturers have also begun offering amplifiers for the L- and S-bands for additional system capacity, though Mulrooney contends that most system vendors are simply tightening the channel spacing within the C-band. While there are specific applications for the L-band today-particularly in South America, Italy, and Japan, where specialty fiber like dispersion-shifted fiber is predominantly deployed-amplification for the S-band will not see any kind of volume for at least the next four years, she asserts.
Though it was once assumed that metro DWDM networks would not need amplifiers, service providers have discovered that for large metro rings, amplification is indeed a necessity. Typical span lengths themselves are below the standard 80 km required for regeneration, but the proliferation of optical components in the metro contributes to signal loss at each node.
According to Mulrooney, for at least the next five years, EDFAs will rule the metro space as well. However, they are not merely scaled-down versions of long-haul EDFAs. Even more than long-haul, the metro area is driven by economics-in some systems, amplification can account for a full 20% of the total cost. Optical-amplifier vendors must reduce the cost of each amplifier by more than half to make them cost-effective for the metro, says Mulrooney.
To lower the cost of metro EDFAs, many vendors are now using lower-cost components, including uncooled pump lasers. Because performance requirements are less stringent in the metro, amplifier manufacturers have begun taking the pump laser out. The wavelengths shift based on the temperature of the pump laser, so omitting the device does affect the overall performance of the amplifier but not enough to make them unsuitable for metro applications.
Moreover, the constant adding, dropping, and reconfiguring of wavelength services in the metro requires a more agile amplifier. "Because you are not necessarily sure what your metro architecture is going to be-how many nodes you'll have-you need amplifiers that are flexible in terms of their ability to set their gain and their output power," contends Jennifer Aspell, vice president of the optical-amplifier group at JDS Uniphase (San Jose, CA).
Variable gain control is particularly important in metro applications, says Marc Schwager, vice president of marketing at Altamar. "Imagine you are adding or dropping 20 wavelengths out of a system or a ring. All the amplifiers need to adjust very rapidly to this disruption; otherwise, the amplifier power will be spread differently across the wavelengths, and it could change the power at any given receiver."
Transient control is also an emerging trend for metro amplifiers, says Aspell; the constant adding and dropping of channels creates transience that must be controlled.
Metro amplifier vendors face the challenge of creating devices that are lower-cost and have a smaller footprint but are no less specialized than their larger, more expensive LH cousins. Among the technologies that may eventually compete with EDFAs in this space are semiconductor optical amplifiers (SOAs), erbium-doped waveguide amplifiers (EDWAs), and-well into the future-maybe even Raman amplifiers.
Though SOAs are generating a buzz among metro players, it will be some time before they are widely deployed-if ever. While they are easier to integrate with other components and offer a lower-cost alternative to EDFAs, SOAs still have critical hurdles to overcome. They suffer from relatively low output power, higher noise, and polarization sensitivity. That said, SOAs may find their niche in single-channel applications to overcome losses in optical switches or crossconnects or to displace EDFAs in single-channel, low-cost scenarios.
CIR's Mulrooney is not convinced, however. "You've got guys like Corning, Agere, and JDS Uniphase saying they can get EDFAs to $1,000 long-term, so that makes it interesting," she says. "If they can actually do that, you wonder if SOAs would ever take the place of EDFAs in that case at all."
Nortel and others are working on an EDWA, which differs from an EDFA in that the fiber is not doped with erbium; the EDWA uses a waveguide design, and the waveguide itself is doped with erbium. While EDWAs have the potential to be smaller and less expensive than other metro alternatives, they also suffer from performance issues in the L-band and are difficult to manufacture, says Mulrooney. "My impression is that those are more lab products at this point," she surmises. "They probably won't see much in the way of sales at least for the next five years or so."
While Raman amplifiers have not yet reached cost parity with EDFAs in LH and ULH, industry insiders are talking about deploying them in the metro. Are they getting ahead of themselves? Mulrooney believes that such deployments are "way far away, if only just because of the cost. Right now, Raman amplifiers are $28,000, and I don't see them getting to where they need to be in the near future.