SOA debate rages on

Touted for their cost-effectiveness, compact footprint, and potential for integration, semiconductor optical amplifiers (SOAs) nevertheless always seem to be two or three years away from being the next big thing. While SOAs are now commercially available from several companies, some industry insiders believe they have not yet made the leap from lab curiosity to network reality.

Thanks to their semiconductor-based design, SOAs can take advantage of the economics of semiconductor manufacturing; in principle, thousands of optical amplifiers can be made on a single wafer, leading to a very compact device. Typical SOAs are a millimeter long and less than half a millimeter wide. That's the upside. The downside is that SOAs are less powerful than EDFAs, which is why they cannot be used in WDM applications that require high output power. There are other applications where SOAs may find their niche, however.

According to Paul May, chief executive of Kamelian (Oxfordshire, UK), there has been increasing interest in using SOAs for reach extension applications. "When you have a link designed for a certain span and you want to extend it from, say, 80 km to 120 km without changing the transceiver modules, you can put in an SOA module to upgrade the link," he explains. "It's pretty cost-competitive compared with actually taking out the transceivers and putting new ones in." While SOAs compete with low-cost EDFAs for this type of application, May believes the SOA's compact size gives it an edge.

SOAs are also finding their way into CWDM networks where low-cost amplification is necessary to extend the capacity and distance of CWDM systems. SOAs are ideally suited for this application, since they are capable of amplification across an extended gain bandwidth, from 1260 to 1620 nm. EDFAs operate only in the C-band (1530–1560 nm) and L-band (1570–1610 nm) and can therefore cover just two CWDM channels at a time.

SOA research and development has been picking up as well, says May, who reports seeing increased interest in the use of the technology for wavelength conversion. The company was initially established to develop wavelength conversion products (hence the name "Kamelian"), but began focusing on SOAs for metro applications when the long-haul market fizzled. "Now we seem to be coming full circle," he notes. Kamelian has developed its own wavelength conversion product, currently in the evaluation stage. The All Optical Wavelength Converter (AOWC) is based on SOA technology and an integrated passive silica waveguide.

Fellow startup Kailight Photonics (Dallas, TX, and Rehovot, Israel) has also developed wavelength conversion/all-optical regeneration products based on SOA technology. The three products in the Tunable All-optical Signal Regeneration (TASR) family provide 2R (re-amplification and reshaping) and 3R (re-amplification, reshaping, and re-synchronization) optical regeneration functions. Because the TASR products are tunable, they can also be used as wavelength converters.

"What is unique about our design is that it uses only a single SOA, which is very low-cost," explains Neil Salisbury, vice president of marketing and business development. "To do this type of regeneration or wavelength conversion today, you have to go back and do an OEO [optical-electrical-optical] conversion. Using a single SOA, we allow you to do the same thing that an OEO transponder would do, yet we can do it all-optically, which means we're fully transparent to bit rate and protocols."

Salisbury believes Kailight's design will provide a huge cost advantage, particularly at 40 Gbits/sec where OEO conversion tends to be very expensive. "At 40 Gbits/sec, going all-optical is going to be much more economical than going OEO," he says. Kailight's modules, the TASR-1010, TASR-1040, and TSAR-4000, are targeted for use in optical crossconnects, switches, routers, regenerators, and reconfigurable optical add/drop multiplexers (ROADMs).

While there's no doubt strides have been made in SOA development, are companies like Kamelian and Kailight getting ahead of themselves by talking about things like wavelength converters, ROADMs, and optical crossconnects? "I'll be the first to say this is not stuff that's going to be commercial in a year or two," admits May, "but it's an indicator that people are starting to think about some of these next-generation systems."

Of course, thinking about next-generation technology is a far cry from actually deploying it. While the darkest days of the telecom crash may be behind us, there's been no indication that all-optical networking will take off anytime soon, contends Tom Hausken, director of optical communication component research at Strategies Unlimited (Mountain View, CA). "The trouble is it's impossible to get inside the mind of the system designer," he says. "Even if you know what they are thinking today, you don't know if six months from now, they'll say, 'Optical crossconnects are definitely the way to go.' These are big decisions, and it's still a very conservative world out there."

JDS Uniphase (San Jose, CA), once considered a key player in the SOA space, agrees with Hausken's assessment of the market. The company abandoned its SOA product line more than a year ago in favor of alternative amplification technologies.

"If you look at what our customers are looking for in terms of performance and price that will enable them to meet carrier requirements now and in the future, it's not a slam dunk for any technology," asserts Dhrupad Trivedi, director of JDS Uniphase's optical-amplifier business. As such, the company decided to take a technology-agnostic approach to the market, examining the pros and cons of SOAs versus erbium-doped waveguide amplifiers and single-channel EDFAs.

For those customers looking at SOAs for booster applications, for example, JDS Uniphase offers a single-channel EDFA that is bigger than an SOA by a factor of three or four but provides much better performance, says Trivedi. "What they are losing in size, they are gaining in performance with price being a wash," he offers.

Going forward, he says, the company will have a replacement product for those seeking arrays of four or eight amplifiers with independent power-level controls. "Instead of going the route of making an SOA array," he explains, "we are going the route of using a waveguide to do the batch processing and integrating that with erbium fiber to get the best performance and processing."

More in DWDM & ROADM