Electroabsorption-modulated laser overcomes performance, packaging, and distance deficits
A novel externally modulated light source design couples laser and modulator parts in a single package to drive 2.5-Gbit/sec signals over fiber-optic cables to beyond 600 km with extreme stability
ray nering and andy zhou
Lucent technologies, microelectronics group
Today`s need to communicate more information faster, and the anticipated exponential growth of bandwidth requirements, have placed increasing demands on existing telecommunications and data communications systems. Much of the anticipated growth is based on explosive Internet traffic in addition to traditional communication needs. These traffic increases have already started to choke the existing capacity of both long-distance and local networks, causing telecommunications bottlenecks around the world.
The need to expand these networks quickly has led service providers to look for ways to upgrade or increase their capacity as quickly as possible. For fiber-optic systems, capacity upgrade can potentially be accomplished in several ways:
More cables could be run. Although this method is being done, it is costly and time-consuming.
The data rate of existing systems could be increased to electronically multiplex more signals together. Currently, most systems operate up to 2.5 Gbits/sec. The ability to increase the data rate over existing fibers to >10 Gbits/sec appears attractive. This approach, however, would require service providers to develop new high-speed terminal equipment and replace existing central office equipment. It also would require that existing optical fiber have the capacity to operate at these speeds, because fiber manufactured years ago is not of the same quality as fiber manufactured today. Due to excessive loss or poor dimensional consistency of the fiber, much of the imbedded base of fiber might not support 10-Gbit/sec transmission over several hundred kilometers.
The third alternative is optically multiplexing signals, primarily at 2.5 Gbits/sec, of various wavelengths over a single fiber, known as dense wavelength-division multiplexing (dwdm). Although not trivial from a network management perspective, this approach allows service providers to up grade route capacity as needed, without requiring more fiber. Therefore, if a link only requires a doubling of capacity, two wavelengths can be operated.
Later, as needs increase, more channels can be added. Some equipment manufacturers can provide systems with up to 40 wavelengths multiplexed on a single fiber. Additionally, the data rates could be different, and even the types of signal could have a different modulation scheme for each wavelength.
Long-haul carriers need to increase the capacity of their systems, but they also want to reduce the number of regenerators. Traditionally, regenerators were located a maximum of 40 km apart for gigabit-transmission systems. Direct-modulated lasers were used to transmit between regenerators. As improved lasers were developed, they were able to transmit over longer distances and bypass some regenerators, thereby saving real estate and site management costs. Network cost and maintenance were also reduced. Direct modulated lasers could be used in such systems for distances up to 250 km.
However, these devices are limited due to chromatic dispersion of the fiber. This dispersion degrades the optical signals because each light pulse is made up of a distribution of frequencies centered around a specified wavelength. These wavelengths have differing transit times through a length of the fiber, and the result is an increasing distortion of the signal as the wavelengths travel further down the fiber.
This distribution of wavelengths is generated by the signal source and is called chirp. To transmit further, for example, in undersea applications, an external modulator can be used. The amount of chirp generated in an externally modulated device is significantly less than that in a directly modulated laser. In the past, this typically meant using a lithium-niobate modulator. Although physically large, sensitive to the polarization of the incoming light, and generally awkward to use, these devices could be used for ultralong-reach applications.
For dwdm systems, two approaches have been taken in the market. The first uses lasers operating in a continuous-wave mode and modulating light with a lithium-niobate modulator. The lasers are selected for wavelength. The modulator allows transmission distances greater than 600 km without regeneration.
The second approach uses an electro-absorptive modulator. Although physically smaller than a lithium-niobate modulator, the device is still polarization-sensitive. It is designed to operate at the same wavelength as the source (unlike lithium-niobate), and the source must be physically connected, typically through a fiber, to the source.
The last approach uses an electro-absorptive modulated laser (EML). In an EML, the laser and the modulator are integrated on a single InP chip. For 2.5-Gbit/sec applications, the EML offers the same performance as a lithium-niobate modulator, but it is smaller and less complicated to use. Unlike the lithium-niobate modulator, the EML is self-contained. The laser can be biased at a constant output power, and the signal applied to the modulator switches the output light on and off.
Under development for several years, EMLs are well-suited for system applications requiring a less bulky package and less cost because the laser and modulator are combined in a single device.
In 1986, the first distributed feedback (DFB) laser that was monolithically integrated with an electroabsorption modulator was demonstrated by researchers using a hybrid liquid phase epitaxy/molecular beam epitaxy growth technique. These early devices were of the butt-coupled variety, in which the laser and modulator are grown in separate steps. However, they suffered from poor coupling efficiency between the laser and modulator, and poor extinction-ratio-performance caused by signal scattering at the abrupt waveguide transition. An important advantage of an EML is that it optically and electrically isolates the laser from internal reflections or electrical crosstalk from the modulator. Reflections can cause the linewidth of the laser section to broaden, resulting in excessive chirp. Crosstalk between the modulator and the laser also causes chirp in the laser. Chirp reduces the maximum distance over which the EML can be effectively operated.
In the early 1990s, the first EML was fabricated with the selective-area-growth (SAG) technique. This method grows the laser and modulator active regions in a single epitaxial step. The results are minimum internal optical reflections and maximum coupling between the modulator and laser. The SAG technique produces high yields and consistently high-performance devices. It is also a simpler and more robust manufacturing process than other methods.
During device fabrication, a trench is etched into the chip surface to provide electrical isolation between the DFB laser and the electroabsorptive modulator section. During operation, separate electrical connections are made to the top of the laser and modulator sections. A common electrical ground is attached underneath both sections. Consequently, the chip and package have been designed to minimize any electrical crosstalk from the modulator (see figure).
The light is generated in the multiquantum-well layers of the DFB section and passed on to the modulator section. The period of the grating effect provides wavelength control for the laser. When a positive current is applied to the laser, it emits a continuous signal down the optical cavity.
The modulator section of an EML is voltage-controlled. Without an applied bias voltage, the EML is transparent to light. When a negative voltage is applied to the modulator section, it turns opaque and absorbs light from the laser section. In this way, a sequence of negative electrical pulses turns the absorption of the modulation on and off, producing a sequence of optical pulses.
When using these devices, the EML requires less than 2.5V of peak-to-peak modulation voltage. Some lithium- niobate modulators require in the range of 7V. To optimize performance, the offset voltage of the modulating signal to the modulator is adjusted. Changing the bias voltage to the modulator affects the chirp of the device. Negative chirp can be generated by more negatively biasing the modulator up to a point. Although this chirp adversely affects the output power and extinction ratio, significant improvement can still be made to the device. Some devices can operate over distances in excess of 1000 km.
The delivery of a high-quality signal is essential in the design of a high-speed, high-performance device. The impe d ance of the signal driver must be well-matched to the device to obtain optimum performance. Although the nominal impedance of the EML is specified at 50 ohms, the actual impedance is about 40 ohms.
The response of the device, however, is far from linear; in fact, the modulator has an S-shaped transfer function. As the EML is modulated, the output pulse has a low crossing point, which produces some distortion in the pulse shape. Typically, this slight distortion is not a problem. However, if a resulting eye pattern is an important design issue, the threshold level of the driver can be adjusted to compensate for the pulse shape.
These laser devices have been available in the laboratory for several years, and their performance has been hampered less by adequate design than by production technologies. Within the past two years, though, significant improvements have been made in EML manufacturing techniques. The EML is manufactured in the same package as the DFB lasers currently used in system applications. This similarity means that the package has been qualified for network use, thereby making the EML a cost-effective, viable option for field deployment. The EML is evolving into a mature device with the volume production and delivery of several thousand devices for dwdm applications (see photo on page 63).
In transmission systems, reliability is critical. When a new device is used in a new application, many performance factors must be investigated to ensure that the system will operate over a long period of time, typically, 20 years. For standard applications--where dwdm is not being employed--the important factors deal with how the device will change over time and whether the various electrical bias loops maintain stable output power, extinction ratio, etc. For their part, dwdm systems have their own set of criteria in addition to those described.
The first criterion for a dwdm system is the selection of wavelengths to precise tolerance in order to fit up to 40 channels in the wavelength range where an optical amplifier operates (1530 to 1560 nm). Typically, this tuning of the EML wavelength is done by changing the temperature of the laser.
Inside the EML package, the device sits on a platform with a thermistor and some optics parts. This platform is mounted on top of a thermo-electric cooler. The cooler electronically controls the temperature of the EML. The wavelength of the EML is somewhat temperature dependent (approximately 0.08 nm/°C). Typically, the EML is tuned to a particular wavelength by adjusting the temperature between 15° and 35°C and using a thermistor for feedback. This design yields an EML wavelength tuning range of approximately 1.6 nm. When a device is shipped, it is usually tested at the wavelength channel of interest. The stability of the tuning electronics must maintain the temperature setpoint within 0.5°C.
Initially, operational dwdm systems tried to space the wavelength channels as far apart as possible. This allowed the use of less-precise dwdm parts and lessened the requirement on any wavelength drifting by the EMLs or the dwdm parts. As the channel counts increased, however, the channel spacing decreased. Smaller channel spacing turned the wavelength drift of the EML into an important design issue.
To that end, some manufacturers have built optical stabilization features into their designs. Other vendors periodically check and re-tune their devices as necessary. Lucent Technologies produces a virtually drift-free EML. The wavelength is set initially at system start-up and is not adjusted for the lifetime of the system (20 years typical).
The industry wants a device that can operate for 20 years at 40 different wavelengths. Our company`s EML is designed to be tuned to a given wavelength within 1 part in 150,000 and which is guaranteed not to change wavelengths by more than 3 parts in 150,000 during 20 years of continuous operation.
Unlike directly modulated lasers, dispersion penalty measurements are typically not needed for EMLs. The performance of a link is highly dependent on the system design, including the number of amplifiers, power in the fiber, and the number of wavelengths transmitted. Lucent has determined that the best way to rate the relative performance of EMLs is by making time-resolved chirp measurements. The chirp of the EML is measured as the device is being modulated. This chirp is generally on the order of tenths of an angstrom (10-10 m). Devices have been developed that have a peak-to-peak chirp for a typical application of less than 0.2 angstroms. With a chirp better than 0.15 angstroms, EMLs can operate beyond 600 km. u
Ray Nering is technical product support manager for optoelectronics products, and Andy Zhou is product manager of electroabsorptive modulated laser products for optoelectronics products, both at Lucent Technologies, Microelectronics Group, Breinigsville, PA.