Optical amplifiers come of age

The market for optical amplifiers should top $7 billion by 2004, while the technologies and applications for amplification will continue to evolve from the basic EDFA to more complex subsystems that also make use of Raman, SOA, and EDWA technologies.

Tom Hausken

A decade since the first erbium-doped fiber amplifier (EDFA) was introduced in 1989, optical amplifiers have achieved phenomenal success, reaching $3.3 billion in sales in 2000. So far, the business has been mainly confined to C-band EDFAs, but that is about to change as products and applications enter a new era of diversification.

Not only are the products and applications expanding, but the emphasis in the industry is changing. While customers once focused on the sheer performance of the latest "gee-whiz" technology, they are now scrutinizing the ability of suppliers to incorporate more functions into amplifier subsystems and quickly ramp up production.

Even the industry structure has changed, as systems manufacturers are off-loading the design and assembly of amplifiers to a new layer of subsystem suppliers. For their part, component suppliers have risen to the challenge by becoming subsystem suppliers, no longer satisfied with only components. Most notable in this category are Agere Systems (formerly Lucent Technologies Microelectronics Group), Corning, E-Tek Dynamics (a division of JDS Uniphase), and SDL Inc. The subsystem suppliers are already moving to the next step by outsourcing the subassembly tasks. JDS Uniphase announced in 2000 that it would be contracting with Toronto-based Celestica to assemble amplifier modules.

Along with this new maturity in the industry comes more emphasis on lower prices, of course. Unit sales of optical amplifiers have now reached a level such that the lower-cost amplifiers become affordable for several new applications. The industry keeps gaining momentum, so that volumes will continue to increase substantially, with the promise that yet more applications will become affordable. Some of these applications are still years away, but many of the more optimistic promises made by suppliers now seem achievable.

Even as the worldwide economy goes through its ups and downs, and as systems manufacturers seek lower-cost solutions, the demand for a diverse array of optical amplifier products becomes more important, boding well for subsystem suppliers. Strategies Unlimited forecasts that the optical amplifier market will exceed $7 billion by 2004 (see Fig. 1), although the revenue will be divided among more diverse technologies and applications than in the past.

Raman becomes the Next Big Thing

The most notable development in the field of optical amplifiers is the sudden emergence of Raman. While Raman was one of the first processes to be studied for optical amplification, dating back to the 1960s, its potential for telecom systems remained at a low profile for years. It remained relatively obscure until the recent rapid expansion of the industry to 10 Gbit/s and now 40-Gbit/s systems made Raman amplification important. Raman amplification is now being designed into new 10-Gbit/s systems, and will be essential for emerging 40-Gbit/s systems. Because of widespread use in long-haul systems and high unit prices, we estimate that Raman sales will reach $3.5 billion by 2004, more than all optical amplifier sales in 2000.

In a curious twist of the technology, suppliers will not actually sell discrete Raman amplifiers, at least not for several years. Rather, suppliers are offering pump blocks that enable Raman amplification in systems (see "What is a Raman pump block?" p. 40). While the concept is simple, the pump blocks are not inexpensive and bring a higher price than EDFA gain blocks. For the time being, Raman pump blocks will be mostly used in conjunction with EDFAs, as a substitute for the EDFA preamplifier stage. One or more midstage access ports allow for insertion of dispersion compensation, gain equalization filtering, or other functions. EDFA stages still will be necessary because the industry prefers incremental changes in design, rather than starting over with a blank sheet of paper.

Suppliers are already suggesting hybrid Raman-EDFA modules that merge the two blocks into one. Almost before it even emerges as a separate product, Raman is becoming incorporated into such new hybrid products. For this reason, we do not see Raman as displacing EDFA sales, but rather as simply evolving into new, more complex products. Such new products will be recognized more for their applications than for the particular technology inside the box.

The elusive metro market

While metro networks will use amplifiers in emerging systems, this market is more complex than commonly characterized. For all the talk about amplifiers in metro networks, relatively few of the vast number of metro links actually use optical amplifiers.

The interoffice segment of metro applications and products sometimes use amplifiers, but those amplifiers are hardly different from those used in many long-haul applications and products. Both carry multiple WDM channels over highly-aggregated backbones. While metro products are adapted to relaxed specifications for dispersion and noise, there is no substitute for performance, even in the metro space.

Conventional EDFA products are evolving in other ways, however, as passive and active gain blocks are transformed into more functional modules. A passive gain block-also called a "glass box"-contains just the essentials of amplification: the specialty fiber, isolators, and couplers. An active gain block also contains the pump lasers.

An amplifier module is essentially a gain block with electronics for control, monitors, and alarms. With the electronics inside, there is little left for the systems manufacturer to add to make it a card or rack-mounted amplifier. The amplifier module is a nearly complete subsystem (see Fig. 2 and Fig. 3).

Manufacture of amplifier modules eases the manufacturing for systems companies. It also allows the supplier to electronically reprogram a common platform, providing more functions and faster time-to-product. These capabilities mean lower cost for amplifier customers-but higher value for subsystem suppliers-compared to passive gain blocks. And higher value means better margins.

Such metamorphoses of one product into another in the amplifier business are common. It makes market forecasting difficult but not impossible. At Strategies Unlimited, we count all units used for amplification, even those made by captive suppliers, as well as components and systems, to obtain the full picture of the industry.

Low-cost amplets

With Raman schemes at the highest performance end of the product spectrum, low-cost amplifiers are emerging at the other end. The industry commonly talks of the need for yet lower cost amplifiers for the metro market, sometimes called amplets. In fact, low-cost amplifiers will be successful in long-haul networks as well, for applications in transparent wavelength switching and 40-Gbit/s systems.

Three technologies are vying for the low-cost amplifier market: low-cost EDFAs, erbium-doped waveguide amplifiers (EDWAs), and semiconductor optical amplifiers (SOAs). All claim they will bring costs down substantially below today`s prices. Substantial unit sales of EDFAs can drive component costs down further. EDWAs claim to be the natural low-cost offspring of the EDFA, using much of the same physics, but in a low-cost manufacturing platform. While SOAs point to laser diode manufacturing as their model for the lowest-cost platform.

SOAs actually represent a long history of development in optical amplifiers. The first reference to optical amplification in semiconductor material dates to 1953, by John von Neumann.1 Indeed, the term laser means "light amplification by stimulated emission of radiation." As soon as laser diodes were successfully developed for telecom applications in the early 1970s, researchers pursued SOAs as well, with the first product appearing in 1989 (see "Timeline of optical amplifiers," p. 38). As it turns out, both SOAs and Raman have long histories, but have had to wait decades for the telecom industry to be able to use them to their best advantage. As EDFAs enter their second decade as a commercial product, the field of optical amplifiers promises to become much more diverse and widespread than in the past, with the help of these long-awaited technologies.

Diverse and promising future

It is not commonly recognized that the EDFA changed the course of optical communications. While EDFAs would be significant because they substitute optical repeaters for expensive electronic regenerators, they became especially important for opening the way for wavelength-division multiplexing (WDM). WDM is cost-effective because the cost of the amplifiers, which remains substantial for long-haul networks, can be spread across many WDM channels. After optical amplifiers enabled WDM, the cost to light a channel became a fraction of what it was before.

The need for greater bandwidth drives the demand for optical amplifiers. Increasing either the number of WDM channels or the data rate increases the need for more amplification. Moreover, today`s systems require more monitoring and remote control of features in the amplifiers, leading to increased complexity of subsystems, which helps keep prices and revenues high, even as the industry matures.

REFERENCE

1. J. Bardeeh, Collected Works of John von Neumann, 5, 420, Pergamon Press, NY (1963).

FIGURE 3. Amplifier subsystems are a level of integration between the pump modules and complete benchtop or rack-mount units.

Timeline of optical amplifiers

It has been said that most technologies take 20 years to become an overnight success. EDFAs seem to have done better than that, with the first developments published simultaneously in 1987 by the University of Southampton and AT&T. Commercialization soon followed and EDFAs have become conventional equipment in long-haul telecom systems.

But even the EDFA has early roots: the first fiber amplifier was demonstrated by researchers at the American Optical Company in 1964. Thirty years after this early work, optical amplifiers came into widespread use in telecom systems. The development of SOAs and Raman have taken longer, with initial efforts also dating to the 1960s. And so ironically, these early technologies studied for optical amplification may turn out to the last to reach maturity (see table).

What is a Raman Pump Block?

With all the talk about Raman amplification, it may surprise some readers to know that no company actually sells a complete Raman amplifier. That is because the gain medium in Raman amplification is the transmission fiber itself. The amplification occurs over several kilometers of fiber already installed in the ground. Such a scheme for amplification is said to be distributed.

In contrast, EDFA products must supply a section of several meters of specialty fiber, which is wound inside the amplifier module. All of the amplification of an EDFA, therefore, occurs inside the module. EDFA schemes are said to be discrete.

Suppliers currently only offer pump subsystems that enable Raman amplification. Strategies Unlimited calls these subsystems Raman pump blocks. The pump block contains, at the least, the pump light source at a wavelength in the 1400- to 1510-nm range, an isolator, and some monitors and feedback controls. More sophisticated designs combine several sources, each at a different wavelength, to obtain a flatter Raman response (see figure).

The pump block can be compared to the EDFA gain block. A gain block includes the gain medium, but like a pump block is only a subsystem. The block must be integrated with other components and equipment to offer full functionality.

Conceptually, the pump block is relatively simple: it is a light source. In practice, however, the subsystem is complex and requires components specially qualified for high optical power levels. The subsystem can cost from $20,000 to $60,000.

Two types of sources are used in Raman pump blocks: laser diodes and fiber lasers. Both produce single-mode light in the 14xx-nm range. The fiber lasers in this range must use a special technique to shift the wavelength from a shorter wavelength to the 14xx-nm range. The wavelength shifting is produced by a specialty fiber known as a Raman cascade. This coincidental use of a Raman technique to obtain the desired pump wavelength is unrelated to the final application.

In a Raman pump block, the output of the pump block is multiplexed to the transmission fiber in the direction counter to the direction of propagation of the signal