Initially, dense wavelength-division multiplexing (DWDM) was simply a means of increasing the capacity of any single-fiber line, providing a path to easy upgrade and expanded capacity. The growing sophistication of DWDM systems has included a shift to higher channel densities, faster data rates, and system flexibility, all of which have affected the design of the lasers at the heart of these systems. Although speed remains the primary driver of laser innovation, the need to integrate more functionality has become a close second.
This trend is evident in the two types of commonly used transmitters: the electroabsorptive modulated laser (EML) and the lithium niobate modulator. Each has its own design benefits and trade-offs.
EML, the workhorse
The EML is a single-chip 1550-nm laser with an integrated electroabsorptive modulator, generally housed in a 14-pin butterfly package. Designed for use in 2.5- and10-Gbit/s extended-reach transmission applications where distances between regenerators are in excess of 600 km, an EML does not require complicated driving electronics. Available in channels across the ITU wavelength grid for DWDM applications, it is the first mass-produced optoelectronic IC-a single chip of indium phosphide with more than one optical component built into it.
Several advantages are inherent in a highly integrated EML. It features a low chirp, compared to a direct-modulated laser, and a low switching voltage. In addition, it eliminates the need for an external lithium niobate modulator that, in the past, required splicing fibers together. Also, the EML is smaller and, in volume, is normally less expensive.
Electroabsorptive modulated lasers were first introduced into the market in the mid-1990s at 2.5 Gbit/s. This year, a 10-Gbit/s EML incorporating an IC to drive it-especially important for impedance matching and the frequency of operation-has been introduced (see Fig. 1).
Lithium niobate, known for speed
Lithium niobate electro-optic modulators achieve exceptionally high bandwidth in multigigabit transmission systems. The modulator device converts the light from a CW laser diode (operating in the 1550-nm range) into a time-varying optical output, corresponding to an electrical input signal. Low-chirp performance results in limited dispersion and long system spans.
A system designer implementing a 10-Gbit/s solution several years ago would have used lithium niobate as the only viable option. Today, lithium niobate devices are available at 20-Gbit/s speeds, with 40-Gbit/s in the offing. When it comes to speed, lithium niobate is most often the technology of choice because it is easily redesigned to achieve higher bandwidth (see Fig. 2).
The trade-off, however, is that switching voltages also increase, providing a challenge depending on the switching voltage and speed of the signal. Lithium niobate devices are also large compared to EML chips, so that fewer devices per wafer are possible during manufacture. Lithium niobate modulators are thus inherently more expensive.
Comparing the two
Both EMLs and lithium niobate modulators are used in the long-haul transmission business because the devices can support transmission over long distances without having to be retimed or sent through a repeater. Amplification is necessary with both types. Selection between the two most often is a function of product availability and features.
Customers, rather than concentrating on long haul for next-generation design, may instead be looking at very-high-data-rate interfaces to terabit routers, a case where lithium niobate modulators are too big and bulky. A customer may opt for the smallest device and lowest cost possible at 10 Gbit/s, yet at 40 Gbit/s want the most-advanced technology possible based on speed-EML and lithium niobate solutions, respectively.
Should a design that makes sense in lithium niobate today not be ready until next year, an indium phosphide device may solve the problem more elegantly, requiring a lower-power source laser because of interim integration. The EML will most likely offer a smaller, highly integrated, and lower-cost alternative.
Another important issue is when to select a speed of 2.5 Gbit/s, 10 Gbit/s, or greater. Customers constrained by a limited quantity of fiber may be interested in higher speeds just to increase the capability of the fiber. In the mid-1990s, there was concern surrounding nonlinear problems or parasitic problems related to fiber speed, in particular polarization-mode dispersion (PMD). It remains a major concern with older fiber but not with new fiber.
The basic concerns of today relate to the data rate that the fiber can support. As new fiber is installed, it is easier to run at high rates. When building a new system the choice is normally to use the current state-of-the-art fiber. If, however, the system is implemented over legacy fiber, many other issues drive the choice of modulator.
Of speed and drive
At a 10-Gbit/s data rate, the total loss of the device is a particularly important param- eter. A laser begins operating with a given milliwatt level of power, but by the time the signal is launched into the transmission fiber, loss has built up from all the components between the laser and the launching point. The components causing the loss could include external modulators, optical multiplexers, or external wavelength stabilizers.
Drive voltage is also a significant concern. Normally, drive voltages for lithium niobate modulators are less than 5 V, compared with no more than 3 V for indium phosphide devices. At 40 Gbit/s, generating a voltage as high as 5 to 8 V becomes a significant issue. Different techniques are under development to run the systems. For example, at 40 Gbit/s, an amplitude modulation technique can be used to generate a return-to-zero (RZ) pulse stream and cut out pulses representing zeros.
In this technique, a two-modulator scheme can be used in which a pulse stream is generated by the first modulator and a second modulator cuts out unwanted pulses. Two modulators in a row, however, cause twice as much loss in this design. Through integration, both modulators can be placed on a single chip of either lithium niobate or indium phosphide. When created in indium phosphide, devices such as amplifiers and lasers can also be integrated.
Packaging integrates functions
Research and development efforts are leading toward higher levels of functionality, not only at the chip level but also in device packaging. Recent advances include integrating an optical locking mechanism along with memory and buffer electronics inside the laser package. Until recently this device was inherently stable in a 100-GHz-spaced system, and wavelength locking was unnecessary. How- ever, active feedback is required to maintain 50-GHz channel spacing while taking into account the wavelength open-loop stability of the laser and the optical multiplexer and demultiplexer. or tunable devices, a locking mechanism that allows the device to find the right channel is also necessary. Another example of higher integration is the ability to take an existing device and specify it to be used over four wavelengths at 50 GHz. This four-channel tunable component features an integrated locker and will most likely be introduced in a transmitter. As time goes on, the number of channels will inevitably increase.
The integration of functionality is not limited to the chip level but also occurs within the optics in the chip packaging. For example, a recent advance is the ability to place within the package electrically erasable memories that can have key data points updated as the devices migrate over time. As technology evolves, the amount of information necessary to use the devices becomes more complicated. For example, to operate a tunable device requires a lookup table containing correct operating points for each channel. Operating over different wavelengths becomes increasingly complicated from a data standpoint alone.
With integrated memory, all the unique information about the device is held inside the device package, including the power level, the voltage necessary to drive that specific power, the data for each wavelength, and guidance on how to get to each wavelength. Potentially, this built-in capability will allow a component to be placed in a transmitter and-as long as the transmitter has the right intelligence-the transmitter will set itself up, review the lookup table, extract the appropriate data, and program the microprocessor running the transmitter. It will be easy to determine whether or not degradation occurred by checking the initial settings and voltage parameters of the device to detect changes.
Other examples of integration are driven by the projected speeds of future systems. As speed increases, it becomes more difficult to match the impedance of the device inside a package. This matching is critical to achieving an adequate optical pulse shape that meets transmission requirements. By integrating a driver inside an EML package, it is possible to achieve excellent performance over a distance of 80 km with very good yields. At 40 Gbit/s, integration of the electronics along with the optical components will be even more critical, requiring optical component and IC vendors to work more closely to develop innovative solutions.
Ray Nering is manager of market development at Lucent Technologies Microelectronics Group, 9999 Hamilton Blvd., Breinigsville, PA 18031-9359. He can be reached at 620-391-2509 or email@example.com.
FIGURE 1. The 1550-nm electroabsorptive modulated laser (EML) is based on an indium phosphide chip and an integrated electroabsorptive modulator packaged with a board containing the drive electronics. It is relatively inexpensive and has been the driver of choice for 2.5-Gbit/s transmission systems.
FIGURE 2. The lithium niobate modulator modulates light from a separate 1550-nm laser diode and has been most frequently used in systems operating at 10 Gbit/s.