OC-192 lithium-niobate modulators spark fiber-optic network capacity
OC-192 lithium-niobate modulators spark fiber-optic network capacity
Serving as a key component of externally modulated lasers for fiber-optic-cable networks, proven lithium-niobate modulators enable cost-efficient, 10-Gbit/sec light transmission; provide outstanding voltage, loss, and extinction parameters; and ensure long lifetime and trouble-free operation
jack lin and paul sanders
uniphase telecommunications products
As network planners rely more heavily on externally modulated lasers for signal transmission in long-haul and backbone applications, high-performance OC-192 lithium-niobate (LN) modulators running at 10 Gbits/sec are expected to play a major role in existing time-division multiplexed (TDM) networks and future high-capacity dense wavelength-division multiplexed (WDM) network upgrades.
Lithium-niobate modulators offer significant advantages over other modulation technologies such as electroabsorptive (EA) modulated lasers, especially in long-span networks (see page 56). For example, LN devices provide higher launch power and negative chirp capability, which places less demand on signal amplification and dispersion compensation. They also offer varied bias control options for greater electronics design flexibility. Moreover, LN modulators allow easier and faster installation via separate laser and modulator sections. All these attributes augment laser-modulator design, installation, and maintenance.
Both TDM and WDM technologies are expected to play leading roles in future optical network upgrades, with deployment determined by network architecture and platform. TDM links operating at higher speed require more-sophisticated electronics and encounter limitations in dispersive and nonlinear effects, but are compatible with conventional network software. The advent of erbium-doped fiber amplifiers has enabled practical WDM systems operating on multiple-wavelength channels in the 1550-nm band, which has led to large-scale deployment plans by the telephone companies. WDM systems are especially attractive in providing an easy upgrade path and potential wavelength routing capability for greater network flexibility. Popular with both technology upgrade techniques is the use of externally modulated continuous-wave (CW) lasers at gigabit transmitter speeds. External modulation of lasers is preferred by many system designers over the directly modulated approach.
Directly modulated lasers suffer from inherent chirping that limits the useful bit rate-distance product of the transmitter, especially when used with installed conventional singlemode fiber. This limitation becomes significant with the new generation of OC-192 (STM-64) transmitters. These products are critical components in TDM transmission and promise to be key elements in high-capacity dense-WDM systems. Consequently, network designers are turning to new OC-192 modulator devices based on proven LN technology.
Two external modulation technologies are currently emerging: LN modulators, which are interconnected to distributed-feedback (DFB) lasers, and EA modulators, which are monolithically integrated with DFB lasers.
The EA approach is intuitively attractive for forming both the laser and modulator devices on the same substrate via a selective epitaxial growth process, with no optical gap between them. Despite this elegant design, though, mating these discrete elements monolithically within tight boundaries is challenging to product manufacturing yield. Achieving device tolerances required to match band-edge operation of the modulator to the lasing wavelength has proven difficult, if not elusive, with existing device manufacturing technology. As a result, the availability of EA-modulated lasers is limited, especially in providing multiple-wavelength transmitter sets along well-defined 100-GHz increments in accordance with the proposed International Telecommunication Union WDM channel-spacing standard.
In contrast, coupling of DFB lasers with LN modulators provides a flexible, modular approach to transmitter design with several benefits (see photo). For example, LN-modulator operation over a relatively broad wavelength range translates into lower laser cost (laser yields at specific WDM wavelengths and output power are much improved over those of monolithic devices).
In addition, the insertion losses of LN devices are favorably offset by their ability to operate with high-power lasers. The higher launch power achieved using LN modulators over that of EA devices allows longer span lengths without the use of expensive amplifiers.
Furthermore, the modular laser design lends itself to serviceability because the DFB laser, which has a more tenuous lifetime than that of the modulator, can be easily replaced (the modulator salvaged) if it fails. In EA devices, the whole unit must be replaced. Transmitter upgrading is straightforward using LN devices as the same laser, and much of the electronics board design can be reused when replacing 2.5-Gbit/sec OC-48 lasers with faster OC-192 devices.
Electro-optic materials such as LN exhibit a refractive index change under an applied voltage. Integrated optical devices that exploit this effect are typically constructed from optical waveguides formed in the crystal production stage by one of two processes: titanium diffusion and annealed proton ex change (APE). Both processes yield wave guides with mode parameters that are compatible with low-loss interconnection to commercial optical fibers.
The APE process produces devices with high stability and single polarization operation, with low loss and high-power capability. Using photolithographic and diffusion techniques, optical wave guides are formed in the crystal production stage. Localized metallization forms the electrodes. An applied electric field changes the refractive index of the waveguide, thereby causing a phase shift to be induced in the optical signal traveling through the wave guide.
Integrated optical devices are formed by monolithically combining straight waveguides, branching elements, and bending segments to form intensity modulators or optical switch arrays. Typical optical switches use directional couplers to achieve the switching function, while intensity modulators are mostly configured as Mach-Zehnder interferometers (MZI). In the MZI configuration, the optical waveguide is split by a Y-branch into two arms. Each arm ac cu mulates a certain amount of phase change through the linear electro-optical effect; then, the phase changes are recombined using another Y-branch to achieve intensity modulation at the device output.
For greater than 1-GHz modulation speed, the commonly used method for gathering phase changes is the traveling-wave modulator scheme. In this scheme, input microwave energy co-propagates along with the optical waveguide a few micrometers around the LN modulator surface. However, the microwave propagation speed is only half that of the optical wave speed in lithium niobate. This characteristic leads to a velocity walk-off that has an adverse effect on the modulation and limits the length of the traveling-wave electrode.
The trade-off in overly shortening the electrode length to counter adverse effects is a decrease in the modulation efficiency, which re quires more drive voltage. In general, an electrode length- bandwidth product of 5-GHz-cm and a voltage-length product of 5V-cm are characteristic of this design. For example, in a 2-cm-long traveling-wave MZI modulator, the modulation bandwidth should be about 2.5 GHz, and the drive voltage should be about 2.5V.
To increase the modulation bandwidth to 10 GHz and beyond, the electrode length must be decreased substantially, which in turn increases the drive voltage. This design approach becomes impractical, however, because most microwave drivers and amplifiers operating at 10 GHz or beyond cannot generate a peak-to-peak output voltage greater than 10V.
In new wide bandwidth modulator designs, though, the velocity-matched modulator is required to operate at 10 GHz or higher speeds. To meet this demand, the microwave propagation velocity is increased by a factor of two to match that of the optical waveguide. With velocity matching, the electrode length can be extended to as long as 3 to 5 cm before encountering the length-bandwidth product limitation. The resulting voltage-length product is increased to useful ranges, typically 15 to 20V-cm, depending on the design and placement of the electrode.
The frequency response roll-off is controlled by the roll-off of the microwaves propagating across the electrode and the launch of micro waves onto the LN; consequently, the longer the electrode becomes, the greater the roll-off effect. These velocity-matched OC-192 modulators feature excellent frequency response and compatibility with commercial drivers.
Biasing and chirp options
Most MZI modulators require a bias voltage to set the device operation at quadrature where the intensity is 3 dB down from the peak of the transfer function. Users have two methods to operate the device at quadrature. In the first method, a small optical signal is tapped from the output signal to get a feedback signal; then, a bias control circuit is designed to stabilize the operation point.
In the second method, a bias-free modulator is passively set at quadrature. Bias-free operation is attractive, in that it requires no bias voltage and associated (additional) control circuits. This permanent method is pre-set at the factory.
For long-haul digital optical communications at OC-192 rates, the fiber- dispersion penalty presents a major hurdle in achieving an acceptable bit-error rate. Transmission over conventional singlemode fiber, which makes up the majority of the installed fiber plant, displays minimum dispersion at about 1300 nm. This fiber plant, however, becomes dispersive at 1550 nm; therefore, different wavelengths (or frequencies) in this band of light travel at slightly different velocities. Even for very narrow-linewidth DFB lasers, differences in velocity cause optical pulses to broaden as they travel through the fiber. In turn, the broadened pulses produce operational difficulties at the receiver where the electrical pulses are reconstructed from the received (broadened) optical pulses.
Without reinstallation of dispersion-modified fiber-optic cables in the link, a digital modulator with an appropriate amount of chirp--that is, a frequency shift in the rising and falling edges of the optical pulse--can markedly reduce the dispersion penalty in the link. Chirp modulators predistort and compensate for a large amount of dispersion compared to conventional zero-chirp modulators. Fortunately, modulators can be produced that cover a range of fixed or variable chirp values. For a MZI modulator with fixed chirp design, the chirp value can still be varied by moving the operation point, thereby providing chirp tunability that can be adjusted to compensate for fiber links of various lengths.
Standard operating wavelengths cover the common lightwave communication bands of either 1550 or 1300 nm (see Fig. 1). For a fixed chirp modulator, the chirp parameter (a) is 0.6 at the quadrature point. A different a in the range of 0.4 to 0.9 is also available. The frequency response S11 and S21 is characterized from 0.13 to 20 GHz. The S21 response is measured from the electrical signal of a calibrated photodetector. Note that the 3-dB bandwidth is greater than 8 GHz. The S11 response is taken from the electrical return loss and exceeds 10 dB in the operation bandwidth.
To generate an operational eye pattern, the modulator is driven with 10-Gbit/sec pseudo-random NRZ (nonreturn-to-zero) pulse streams and an OC-192 driver operating at ۭ.2V (see Fig. 2). The modulated optical signal is fed into a Bessel filter to cut off the high-frequency components at the receiver end. The crossing point is at 49%, and the RMS jitter is 2.4 psec. The fall and rise times are limited by the filter. Without the filter, they fall in the range of 15 psec. u
Jack P. Lin is product line manager and directs 10-Gbit/sec and high-speed modulator products, and Paul Sanders is business development manager, new product sales and marketing, at the electro-optics division of Uniphase Telecommunications Products, Bloomfield, CT.