Optical-amplifier technology is evolving to address a wider spectrum of applications and new markets.
James Jungjohann and Rick Schafer, CIBC World Markets Corp.
The development of erbium-doped fiber amplifiers (EDFAs), more than any other technology, has enabled the proliferation of high-bandwidth data networking. EDFAs are a major advancement in photonics, in that these devices sharply reduce the need for signal regeneration, a process by which a photon is converted into an electron, amplified, then converted back into a photon. An EDFA is essentially a special piece of optical fiber that is lined or "doped" with the rare earth element erbium. Erbium, when excited, amplifies light in the low-loss 1,550-nm spectral range used in optical-signal transmission. Optical 980-nm pump lasers, from JDS Uniphase and SDL, transfer high levels of energy to the special fiber, energizing the erbium ions and boosting the optical signal(s) passing through.
This ability to simultaneously amplify multiple wavelengths or signals regardless of bit rate makes EDFAs a critical component of dense wavelength-division multiplexing (DWDM) systems. An important milestone in optical networking occurred in 1993 when MCI announced a successful fiber link between Sacramento, CA, and Chicago, using EDFAs powered by 980-nm pump lasers. (Previously, only 1,480-nm pump lasers based on mature indium gallium arsenide phosphorus (InGaAsP) material systems, were in service, but these devices were limited in both bandwidth and range.) Other service providers quickly followed suit and 980-nm pump demand burgeoned, soon eclipsing the 1,480-nm pump laser in market share. Today, optical signals can travel more than 100 km between amplifiers.
Meanwhile, WDM systems engineers continue to design systems with higher channel counts and increased transmission speeds. At the same time, the push is on to develop technologies utilizing more of optical fiber's inherent bandwidth. Systems designers are thus creating WDM architectures that use the space outside the conventional (C)-band, incorporating both the long (L)- and short (S)-bands to transmit signals. All of these factors are driving demand for increasingly powerful optical amplifiers.
To achieve that, EDFA engineers have begun to design amplifiers with higher power pump lasers and additional pumping stages. A typical high-power EDFA today could have three or more stages in contrast to one when the EDFA was first introduced. By 2004, we expect average pumps per high-powered amplifier to more than double to about 3.5 980-nm pumps and 4.5 1,480-nm pumps. Because amplifier design gets too complex beyond eight pumps, we expect higher- powered chips, providing more milliwatts per module, to play an increasingly important role in boosting amplifier power.
Many large, original equipment manufacturers (OEMs) compete in the fast-growing EDFA market, including Corning (the largest supplier), Lucent Technologies, JDS Uniphase, Nortel Networks, Alcatel, and Pirelli. Notably, JDS Uniphase is now uniquely positioned (with the recent addition of 90-nm pump packaging capability through its recent SIFAM acquisition) as a completely vertically integrated EDFA manufacturer. The total EDFA market could reach $9.6 billion by 2004, from $1.3 billion in 1999, according to our estimates.
Raman pumping, recently introduced by SDL (for remote pumping) as well as Lucent, looks encouraging. The addressable market for remote pumping is projected to reach $200 million within five years. Although widespread terrestrial deployment is not expected for a couple of years, the total market for Raman amplification could grow to $750 million by 2004, from about $3.3 million in 1999.
An alternative method for amplifying optical signals is the semiconductor optical amplifier (SOA), which is being pursued by JDS Uniphase and others. We do not expect the SOA market to develop for several years as suppliers work on introducing a commercially deployable device. We estimate the market could begin taking off in 2001, ramping to nearly $200 million by 2004.
Although the history of Raman scattering effects in optics dates from 1928, distributed Raman amplification, which increases the distance between optical amplifiers, may play an increasing role in high-speed optical networks above the current 10-Gbit/sec (OC-192) to 40-Gbit/sec (OC-768) transmission rates.
In current versions, optical amplifiers are enhanced by Raman amplification in which high-power laser light is sent in the direction opposite that traveled by the source signal. The entire length of the fiber itself is acting as an amplifier. The effect is an almost simultaneous, pushing and pulling of the optical signal. The signal is strongest at the end receiver portion of the network and weakens as it moves toward the source of the signal resulting in a zero to negative noise factor (versus the 4 dB to 5 dB of noise typically added by traditional 980-nm/1,480-nm pumping). With this technology, crosstalk, a typical problem of WDM amplification, is minimized. We see Raman amplification as a complement to, not a replacement for, traditional 980-nm/1,480-nm EDFA amplification.
Today, a key application is for remote pumping (amplification) of distances between 300 km and 400 km, making it appropriate for long-haul and undersea networks. Distances of less than 300 km do not require amplification. A source at JDS Uniphase says some 2,000 to 3,000 such systems could go to bid each year over the next five years. Currently, JDS Uniphase has Raman pumps in the lab and has not yet released commercial versions. Lucent showed a Raman amplified system at Telecom 99 in Geneva, Switzerland in October 1999. SDL has announced several Raman pump versions and is currently shipping to one customer for undersea festooning applications and is in trials with ten additional customers.
To get the necessary powers out (1.5 W-2 W), Raman pumps need about 10 W of multimode power input. Currently, SDL stacks 10 high-powered 800-mW broad area 920-nm modules to get 1.5 W-2 W of 1,480-nm power output. In the future, more powerful lasers (2 W) could reduce the laser count to five per device and bring down costs to achieve mass commercialization. Raman cost reductions are especially important because the alternative likely will involve stacking 10 relatively cheap 140-mW 1,480-nm lasers in series (cheap, brute force) to get similar output.
Raman amplifiers should be commercially successful. Raman gain has several attractive features, including a broad spectral range. New fiber deployment such as Lucent's AllWave fiber offers another application for Raman pumping. AllWave fiber operates above and below the standard 1,550-nm erbium band. Current EDFAs can operate only within the erbium window, making Raman amplification the only way to pump the 1,292-nm to 1,660-nm spectrum.
Raman amplification can also be used in any type of optical fiber. This may be why Lucent is so focused on Raman commercialization as many carriers' installed base of fiber (e.g., AT&T's) is sub-optimized for WDM.
Expect increasing use of Raman technology in remote amplification areas. Remote preamplification delivers the largest improvement in performance at more than 7 dB. Furthermore, re trofitting current WDM networks with Raman amplification appears easy, offering carriers an upgrade path.
Another reason for Raman amplification's attractiveness is cost. In a Telcordia (formerly Bell Labs) trial in mid-1999, scientists transmitted a 40-channel, 40-Gbit/sec signal a distance of 400 km, with Raman amplifiers spaced every 100 km-farther apart than the typical EDFA spacings in high-channel count systems.
An alternative method for amplifying optical signals is the SOA technology pursued by JDS Uniphase and others. The SOAs will be packaged in laser- diode-type butterfly modules, offering operators substantial cost savings. These cost advantages, coupled with the SOA's ability to amplify across a broader spectral range (like Raman pumping), should appeal to metro-WDM system designers. However, we believe that while the current generation of SOAs is improving, they remain too noisy (low signal-to-noise ratio) and need to improve output powers to accommodate new high-speed, high-channel-count, short-haul WDM systems. We do not believe SOA powers will be able to support long-haul transmission; in fact, there is some question as to whether SOA can achieve powers great enough to operate effectively at standard 2.5-Gbit/sec metro speeds. Low power (<14 dB), low-cost (sub-$2,000) EDFAs using cheap pump lasers and very small spools of erbium-doped fiber could present the greatest challenge to SOAs in the metro market.
Another potentially interesting application for SOAs is dynamic wavelength conversion. The SOA takes the incoming photon, shifts its frequency (e.g., from a blue wavelength to a green) and outputs a new wavelength-amplifying the signal in the process. This ability to tune selectively across a wide 30-nm to 40-nm spectral window allows the SOA to function much like a tunable filter-with added amplification.
Despite reports to the contrary, optical amplification will indeed play a large role in metro architectures. While the long-haul WDM market uses amplification primarily to make up for absorption losses; short-haul WDM will depend on optical amplifiers to make up losses from routing, multiplexing, and switching-all of which result in signal loss.
WDM technology's high cost, however, is hampering its acceptance in metro applications. Component suppliers have felt compelled to develop lower-cost technologies. Thus, a low-cost optical amplifier, known as an "amplet" could be introduced as soon as this year with an initial price tag around $5,000.
Amplets, probably single-pumped EDFAs, would sit on the outputs of switches and optical add/drop multiplexers to rebalance powers. The pumping power would be provided by a low-cost, lower-power, 980-nm pump laser. These lasers would not compete with the cutting-edge pumps, which will have incredibly tight tolerances, but instead, will play in a market with looser specifications where capacity is not much of a concern. The pump modules most likely would sell for less than $500 versus the roughly $1,800 price tag for the higher-powered modules.
With JDS Uniphase and SDL basically doubling chip capacity, overall capacity in 980-nm pumps could soon be much less of an issue. In addition, this lower end market will be more inviting to competition, with Spectracom (now a division of ADC Telecommunications), Lasertron (now a division of Corning), Furukawa, and others likely to find a larger market for their products. Given its single-pump design and lower specifications, the amplet could eventually sell for less than $2,000.
In addition to being attractive to the regional Bell operating companies (RBOCs), the amplet could hold mass appeal in the cable-TV market. Working to offer two-way communications over the cable network, AT&T has embraced WDM as key to its hybrid fiber/coaxial (HFC) architecture. As a part of this new design, we expect to begin seeing low- cost EDFAs (amplets) pushed deeper into the network.
Despite all the advances in componentry, we expect average distances between EDFAs to keep dropping. At the same time, prices should fall slowly over the next few years from around $20,000 today to $15,000 by the end of 2004, supporting a 46% compound annual growth rate (CAGR) in dollars over that time.
James Jungjohann and Rick Schafer are equity research analysts at CIBC World Markets Corp. in New York.