EMLs drive WDM applications, offer higher speed potential


Erin Byrne

Electroabsorptive modulated lasers have become established transmitters for 2.5- and 10-Gbit/s transmission. Reaching 40 Gbit/s will require better coupling between laser and modulator, tighter matching of impedances, and further integration.

Electroabsorptive modulated lasers (EMLs) combine a distributed-feedback laser and an electroabsorption (EA) modulator, both in indium phosphide (InP). Introduced six years ago, EMLs are versatile, operating at both 2.5- and 10-Gbit/s, in either the 1.3- and 1.5-µm range, at fixed wavelengths, or a choice of tunable ones. Given this flexibility, they are becoming fixtures in WDM networks. The industry' s first high-power tunable 2.5- and 10-Gbit/s transmitters are based on EMLs, and the technology is being developed for 40-Gbit/s transmission.

Despite this history, plenty of challenges remain to prepare EMLs for higher-bandwidth networks: multiplying their speed many times over, reducing loss, cutting down on parasitic capacitance, mating the RF characteristics very tightly with other devices in the transmitter, and further integration, both on the chip and in the transmitter module.

Single-wavelength EMLs offer price and performance in between that of directly modulated lasers and continuous-wave (cw) lasers with external modulators that are typically made of lithium niobate (LiNbO3). Depending on the application, designers can optimize EMLs to extend reach, increase bit rate, and lower costs.

Monolithic, single-chip fabrication of the EML results in tight coupling between laser and modulator: light is carried from the laser to the modulator through a self-aligned waveguide that is designed into the chip (see Fig. 1). The modulator operates between transparency and absorption, depending on the applied voltage. In addition, the EML isolates the laser electrically from the modulator, reducing electrical crosstalk, and careful design can minimize internal reflections generated by the modulator. These reflections can result in broadening the laser linewidth, increasing signal chirp and thus decreasing reach.

Form factors of the EML match standard slots on circuit boards, although the devices replace separate lasers and modulators that normally occupy a slot each. The 2.5-Gbit/s module comes in a standard 14-pin butterfly housing. The 10-Gbit/s module has seven pins on one side and an RF connector on the other. Lower costs result from combining two optical devices on the same chip, meaning EMLs are less expensive to manufacture than separate laser and modulator chips. They are also less burdensome to store and install than multiple components.

For now, the primary use of EMLs is for metro and long-haul 2.5-Gbit/s networks, and for metro 10-Gbit/s networks. Single EMLs operating at 2.5 Gbit/s have demonstrated transmissions of up to 1000 km, generally with a boost from one or more erbium-doped fiber amplifiers. At 10 Gbit/s, their longest current reach is about 80 km, which is beyond the information transform limit of 55 km. Links up to 80 km are possible without optical amplification using an avalanche photodiode receiver.

Designers can reduce chirp relative to directly modulated lasers to optimize high-speed EMLs for metro use. Since EMLs have inherently low chirp, expensive bit-error-rate testing is not necessary for metro 2.5-Gbit/s applications. The tradeoffs are savings in cost and shorter lead times to volume manufacturing and delivery, useful since metro networks need many transmitters. Chirp in a 2.5-Gbit/s EML can be very low—for example, 0.025 nm from one wave peak to the next.

Designers can vary the EML inputs to influence factors such as signal power or extinction. Increasing either extends reach. Higher power increases the distance the signal will travel, and a higher extinction ratio yields more clearly defined pulses and eases the receiver' s task. A higher extinction ratio also results in fewer missed bits. The choice of which to maximize depends on network architecture and other components used.

To reach higher speeds and longer ranges, EML development aims for even greater integration—building more devices into monolithic chips during fabrication and packing more discrete components into transmission modules—as well as increased bit rates and reduced chirp. One example of integration is locating the IC driver, which provides the voltage to swing the modulator between on and off, as close to the modulator as possible. At 10 Gbit/s, drivers are often gallium arsenide, so they cannot be grown on the same epitaxial chip as the laser and modulator, but they can be very tightly coupled within the same package.

Such a hybrid module optimizes RF characteristics of the transmitter combination, such as permitting a closer match between driver and modulator impedances. The integration also minimizes inductance, which increases with distance and degrades the impedance match. Better matching yields better bandwidth through lower return loss. A few millimeters are far superior to a centimeter or more. Integration of the driver doubles the range of the EML from 40 to 80 km at 10 Gbit/s.

Another major factor that increases bandwidth is careful design that eliminates sources of parasitic capacitance, which is a drag on transmission speed. Coming next are integrated wavelength lockers, which are beneficial in any dense WDM system. The lockers will be necessary for lambda channels spaced at 50 GHz or closer. Integrating the locker into the EML transmitter package reduces signal loss between components, and saves cost and real estate with a single package rather than multiple units.

A few issues remain in realizing this integration. Electronics to control an internal locker need to be supplied by the customer, just as they would be for an external locker. In addition, integrated lockers need to work flawlessly over the entire case temperature range specified by the system integrator, as well as over the lifetime of the laser. Addressing these final design issues will help displace external wavelength stabilizers.

Most InP lasers, including tunable ones, integrate well with EML technology. Many tunable lasers use distributed Bragg reflectors (DBR) to reflect the signal. One challenge in incorporating a tunable DBR into an EML module is providing a reliable control system to optimize performance at each wavelength channel and to compensate for laser aging over time.

One type of EML in 2.5- and 10-Gbit/s tunable transmitters combines five devices—a gain section, DBR mirror, optical amplifier, power monitor, and modulator in a single chip (see Fig. 2). The gain section provides power into the DBR, and the optical amplifier boosts the signal between the laser and modulator to compensate for loss in the modulator. Relative to a single-wavelength EML, the tunable version requires an onboard amplifier because of optical reflection and absorption losses in the tuning mirror. The power feedback loop goes to the amplifier section for two reasons: first, the amplifier section is operated in saturation mode, which means that if the gain-section current is increased, the amplifier realizes only a small increase in output power; second, changes in the gain current affect transmission performance more than amplifier-current changes.

The 2.5- and 10-Gbit/s EMLs can fit into tunable transmitter modules that include modulator drivers and wavelength lockers guaranteeing lifetime stability of ±20 pm. They offer considerable range as well. The 2.5-Gbit/s transmitter, for instance, offers 360-km reach over one of 20 individual 50-GHz wavelengths.

For high speed and long reach, cw lasers with external LiNbO3 modulators are currently the transmitter of choice for two reasons: chirp in their signals is more easily tuned and thus minimized; and they can achieve higher extinction ratios, making rapid-fire signals easier for receivers to read. Lithium niobate transmitters, which modulate by interferometry, can currently send 40-Gbit/s signals a few kilometers over standard fiber. However, linear pulse compression and other dispersion management techniques can extend the reach to more than a thousand kilometers.

Most available 40-Gbit/s solutions are based on external, standalone modulators. Making an EML suitable for 40-Gbit/s transmission demands an extremely agile modulator. Also, many customers want 40-Gbit/s signals in the return-to-zero (RZ) format, which inserts a zero after each information-carrying pulse, rather than the nonreturn-to-zero (NRZ) format typical of lower bit rates, so the transmitter must have even higher bandwidth.

Work is currently under way to move EMLs into the 40-Gbit/s sphere. Paired modulators can double-time the bit rate—one modulator to carve a sine wave from the emitted light, and one to encode the signal by selecting which pulses to transmit. The first 40-Gbit/s systems deployed are expected to use this technique. Although most embodiments of this technology use LiNbO3, an EA version of the tandem modulator approach has been demonstrated for long-distance transmission.

Both modulators are made in InP using EA technology, with the usual benefits of tight coupling of devices and lower cost (see Fig. 3). After this technique is proven, the device might evolve to include the laser and other components now becoming standard on lower-bit-rate chips. The EML could then be packaged in a hybrid-integrated module containing still other components for an easy-to-use, cost-effective transmitter.

Erin Byrne is senior manager for product marketing at Agere Systems, 9999 Hamilton Blvd. Room 1B-142R, Breinigsville, PA 18031. She can be reached at ebyrne@agere.com.

More in Electronics