Semiconductor optical amplifiers expand commercial opportunities

LONG A SUBJECT OF RESEARCH, SOAS ARE FINDING NUMEROUS COMMERCIAL USES. LED BY APPLICATIONS IN OPTICAL SWITCHING AND INSTRUMENTATION, SALES SHOULD RISE FROM $48 MILLION IN 2000 TO MORE THAN $900 MILLION IN 2010.

Jeff D. Montgomery, Stephen Montgomery, and Saba Hailu

The global consumption of semiconductor optical amplifiers (SOAs) will expand rapidly, from a modest $48 million in 2000 to $903 million by 2010. The primary near-term use of these devices will be as switching elements in all-optical switches and optical crossconnects, and in WDM links and other digital fiberoptic transmission applications.

The telecommunications segment of SOA consumption will reach a 48% share of the total market or $125 million by 2005 (see Fig. 1). Laboratory field-test and built-in instrumentation also will be a major SOA user, with 38% value share in 2000, declining to a 15% share or $133 million, by 2010. The leading long-term application of SOAs in the 2005 to 2010 period will be as optical line amplifiers, with a 54% value share in 2000, decreasing to 43% or $388 million by 2010 (see table). Wavelength converters using SOAs will become significant beyond 2001.

ALL-OPTICAL SWITCHES
Photonic switch elements represented a major share of the global consumption of SOAs in 2000, with a 42% value share. Some switch architectures also use SOAs for amplification only; these are counted as switch SOAs by ElectroniCast. Also, some SOA switches use a number of SOAs integrated into a single substrate. In a previous ElectroniCast study, each SOA in these optoelectronic ICs is counted as one device.

Over the forecast period (2000 to 2010), the all-optical switch element share will decline slightly to 41% in 2005, and remain at 41% while expanding to $370 million by 2010. The use of SOAs as wavelength converters was only at the laboratory-evaluation level in 1998, but will expand to a 16% share or $145 million by 2010.

Semiconductor optical amplifiers are similar in construction to semiconductor lasers. They consist of a gain (active) section and a passive section constructed of a semiconductor material such as indium phosphide (InP; see Fig. 2). The main difference is that SOAs are made with layers of antireflection coatings to prevent light from reflecting back into the circuit. Optical gain occurs as excited electrons in the semiconductor material are stimulated by incoming light signals; when current is applied across the p-n junction the process causes the photons to replicate, producing signal gain. The gain medium can be either a bulk or a multiple-quantum-well active layer.

Semiconductor optical amplifiers are attractive for broadband gain, capable of integration with other devices, and offer potentially lower cost than fiber optical amplifiers in certain situations. Significant use of SOAs is anticipated for metro and access applications. SOAs from Alcatel, Kamelian, and OptoSpeed have been commercially available for the last few years, while such companies as Genoa have announced that they will make their unique chips commercially available in 2002.

EUROPEAN R&D LEAD
Much of the European SOA development efforts have been encouraged by European Community ESPRIT and RACE program funding, focused on telecommunications applications in the early to mid-1990s. Substantial SOA work continued at Alcatel, and SOA components and subsystems are part of the company's product offerings. Other European developers included Uniphase/Philips, which was later acquired by JDS; Siemens; and Ericsson. More recent startup European companies include Optospeed in Switzerland and Kamelian in Scotland.

North American SOA development was strongly supported by the US Defense Advanced Research Projects Agency, aimed mainly at ultimate use in defense-oriented fiberoptic links, satellite lightwave communication, and optical computers. Lucent Technologies Bell Labs continues to lead SOA R&D efforts. Within the last couple of years some new North American companies pursuing SOA components and subsystems have emerged, including Genoa, Agility, and Axon.

Japanese SOA development has been aimed mainly at remaining current in semiconductor technology rather than developing commercial product. Some of the Japanese developers included NEC, Mitsubishi, and NTT. NTT also is developing SOAs for packet switching; they have reported >30-dB gain with 6-dB noise figure.

OPTICAL SWITCHING
To handle the increased fiber throughput of DWDM systems, line rates exceeding terabit levels are becoming a requirement. In typical optical networks, crossconnects operate by converting optical signals to electrical signals and then converting them back to optical signals. Optical networks are evolving to use crossconnects containing an all-optical switch, which will reduce or eliminate the OEO conversions. This evolution will allow wavelengths to be managed at the optical level to decrease the cost and complications of electrical devices. Network operators will be able to add or remove wavelengths with the help of reconfigurable optical add/drop multiplexing units.

Semiconductor-optical-amplifier gate switching is one of several technologies being explored and implemented in optical crossconnect switch fabrics. Researchers have reported fabricating monolithically integrated, low-loss switch matrices containing several SOA gates (see Fig. 3). Hybrid integration of SOA gate arrays and silica-based planar lightwave circuits is being pursued by companies including Alcatel, NTT, Kamelian, and Lucent.

WAVELENGTH CONVERSION
Dense wavelength-division multiplexing has brought with it the fact that a single fiber is the conduit for dozens to hundreds of individual signals. This capability is a significant change from a system where there is a one-to-one correlation between the fiber and the signal it carries. DWDM means that optical switching in its most basic form is the switching of wavelengths within a fiber, not the wholesale simultaneous switching of all signals within a fiber.

At some point in the data transfer system, it will be necessary to extract and route individual wavelengths within a fiber. Whether this function is accomplished by means of a switch element and a separate wavelength-selective element or by means of a switch with such properties, the function is still wavelength-selective switching—the need for which is an inevitable consequence of the adoption of DWDM.

In the short term, the first applications for these products will be add/drop and bypass switching of single-frequency, single-band, or comb-structure interleaved sub-bands related to the varying wavelength spacing of the different DWDM networks. SOA-based frequency-selective switches are commercially available as qualified telecommunication products from such companies as Alcatel. The SOA switch has an added capability of being suited to integration with advanced features such as amplification and wavelength translation.

The major advantages of SOA as an optical switching element include very fast switching (1 ns); gain compensation for intrinsic loss; high reliability in severe environment; higher-level integration potential; compact size; and potentially lower cost. Disadvantages include relatively high noise figure; susceptibility to electrostatic shock-induced failure; very high cost of early-production devices; and narrow band amplification.

NOT FOR THE LONG HAUL
Semiconductor optical amplifiers are not an ideal solution for long-haul multiple DWDM channels. WDM amplifiers must be broadband with uniform gain and low noise. The crosstalk between wavelengths in SOAs is excessive—even at the widest ITU spacing (1.6 nm or 200 GHz)—and is an inherent problem caused by the short lifetime of the carriers in the semiconductor materials. Carrier lifetime is only a few nanoseconds, similar to the period of two beating signals separated by typical ITU grid spacings. These beat signals modulate the carriers and generate sidebands, which constitutes four-wave mixing. The mixing problem becomes more severe when the SOA is run in saturation mode, as is typical. In the erbium-doped fiber amplifier (EDFA), in contrast, the lifetime of the erbium ions is much longer—perhaps 10 ms—so they are much less susceptible to modulation effects.

Crosstalk between wavelengths in a WDM system can be a significant problem for SOAs. A major factor in the magnitude of crosstalk is the level of saturation of the amplifier output power. Crosstalk can be substantially reduced by splitting the input power and dividing it among a number of SOAs, then recombining the outputs of the SOAs. However, this results in low optical signal-to-noise ratio. Whether this ratio is economically feasible depends upon the cost of the individual SOAs, and the cost of fabricating them in an integrated assembly.

AS LINE AMPLIFIERs
The advantages of the SOA as a line amplifier include the reliability, compactness, and integratability noted earlier, plus high power output. There also has been the perception that with the expected high-volume production of SOAs, price will drop precipitously. Unlike optical fiber amplifiers, the SOA is available at wavelengths from 600 to 2000 nm.

As a line amplifier, the SOA faces competition from the EDFA in the 1500- to 1600-nm band, and from praseodymium-doped fiber amplifiers (PDFAs) and Raman amplifiers in the 1300- to 1700-nm band. In the 1500-nm band in particular, the EDFA has much higher gain than the SOA (30- to 40-dB small signal gain, double-pumped, versus <30 dB), lower noise, and the acceptance advantage of several years of high-reliability performance in substantial quantities in optical networks. With maturity, volume production, and strong competition, the 1500- to-1700-nm-band EDFA gain-block price has dropped rapidly over the past four years and will continue to decline.

In the 1300- to 1500-nm band, the competitive threat from fiber amplifiers is much less. The PDFA has not achieved high-volume production, and therefore is now more expensive than the EDFA. Also, the power-conversion efficiency of the PDFA is much lower than the SOA. The SOA line amplifier has its strongest potential in the 1300- to 1500-nm band for optical communication. In the long term, the SOA is projected to provide much higher power output, with better power-conversion efficiency, compared to optical fiber amplifiers. In general, linear WDM systems based on SOAs have moderate capacities and short reach. However, their performance and potential cost advantage may be adequate for metro applications.

Jeff D. Montgomery is the founder and chairman, Stephen Montgomery is the president, and Saba Hailu is a senior research analyst at ElectroniCast, 800 South Claremont, Suite 105, San Mateo, CA 94402. For more information, contact Stephen Montgomery at 650-343-1398 or Stephen@electronicast.com.