Overcoming fiber impairments as industry moves to 40 Gbits/sec
Optical signal processing technology eliminates operational issues associated with compensating chromatic and polarization-mode dispersion at high bit rates.
LAURA ADAMS, JONATHAN KING, and SUDEEP BHOJA, Big Bear Networks
Moving the optical industry to higher data rates raises significant network deployment and operational issues for carriers and service providers. As transmission speeds and fiber distances increase, fiber impairments such as chromatic dispersion (CD) and polarization-mode dispersion (PMD) become far more severe, presenting barriers to cost-effective system deployment and ongoing operation.
Addressing these challenges in a simple and economical manner will require next-generation communications equipment to incorporate intelligent, highly integrated electro-optic interface assemblies. Such assemblies will need to take a closed-loop system approach-with optical and electrical components being designed together with a solid understanding of the end-to-end fiber channel-and exploit advanced signal processing techniques. An emerging market known as optoelectronic signal processing (OSP) closely couples photonics and electronics with signal processing capabilities to enable simple economical plug-and-play solutions for high-speed optical links.
For high-speed applications, optical transmission impairments place stringent requirements on the optoelectronic components for the interface-even for short-reach links (~2 km) at 40 Gbits/sec-demanding a closed-loop system design. End-to-end system simulation is needed to define required component specifications. For example, acceptable dispersion penalty for the optical link dictates chirp requirements for the optical modulator, but optimizing chirp may affect other parameters such as extinction ratio, insertion loss, and transmitter patterning.
Transmit signal characteristics are determined by the interaction of the optical modulator with electronic multiplexer and amplifier components in the transmit chain-and all components have to be specified in relation to one another. To satisfy end-to-end link performance, multiple system parameters must be optimized collectively, with optoelectronic interface solutions designed with consideration for the fiber channel. This closed-loop system approach becomes increasingly critical as extended fiber distances cause more severe impairments.
Profitable service delivery over a fiber-optic network hinges on reducing both the operational complexities and large capital expenditures associated with high-speed optical transmission over distance. Two signal distortion mechanisms that raise economic barriers to high-speed optical link deployment are CD and PMD. Finding economical solutions to compensate for these impairments via closed-loop system design is critical to reduce 10- and 40-Gbit/sec network costs for commercialization.
Beyond dispersion, high-speed networks also face challenges posed by optical signal-to-noise-ratio (SNR) requirements and fiber nonlinearity. As the bit rate of optical transmission increases, more signal power is required at the receiver to maintain the same error performance for a given fiber distance. But simply turning up transmitter power is not the answer, because fiber nonlinearity increases signal distortion as optical power rises.
At high bit rates, self-phase-modulation (SPM) becomes the dominant nonlinear effect leading to system penalty. Some of the electronic signal processing techniques under development to address dispersion may also be applicable in partially mitigating penalties associated with other intersymbol interference effects such as SPM.
CD causes different wavelength components in a light pulse to travel at different speeds through fiber, spreading out the optical signal (see Figure 1). The extent of the pulse spreading increases with fiber distance. Adjacent light pulses begin to overlap in time, resulting in detection errors at the receiver.
This CD effect worsens at higher bit rates for two reasons. First, as the rate increases, the bit interval duration decreases proportionately, causing a pulse to spread into the next bit more quickly. Second, when data is modulated onto laser light, the spectral width of the resulting light pulse is broadened to the modulation bandwidth. Higher modulation rates result in a wider wavelength spectrum for the pulse.
For these reasons, CD effects increase as the square of the bit rate, with effects at 40 Gbits/sec becoming 16 times more severe than at 10 Gbits/sec and 256 times more severe than at 2.5 Gbits/sec. At 10 Gbits/sec, dispersion compensation is needed for fiber links beyond about 50 km, whereas 40 Gbits/sec requires compensation beyond about 3 km. Figure 2a shows a received 40-Gbit/sec signal after transmission through 10 km of standard singlemode fiber (SMF). Severe eye closure and distortion are visible, resulting from the effect of CD.
PMD can occur because light travels in two orthogonal polarization modes in fiber. Random deviations of the fiber symmetry cause light in each polarization to travel at different speeds (see Figure 3), leading to pulse spreading over distance and resulting in detection errors. The delay between the two polarization modes is referred to as the differential group delay (DGD).
Beyond first-order DGD, higher-order PMD effects, including polarization-dependent CD and rotation of the principal states of polarization, may induce additional penalty. The total delay between polarizations varies with time as the fiber's physical characteristics change (e.g., due to temperature or mechanical disturbance.) These fluctuations with temperature and vibrations can be dramatic, making PMD an even more insidious problem than CD.
Compensation requires an automatic tracking scheme that can follow the fiber PMD statistics, which change on a millisecond time scale. Although PMD effects increase linearly with transmission speed, the resulting system penalty increases as the square of the bit rate. At 10 Gbits/sec, PMD is typically a problem only for long-haul systems, while at 40 Gbits/sec, PMD can become a significant issue for metro/regional systems as well. Figure 4 shows a typical 40-Gbit/sec PMD corrupted signal with a full bit period of DGD.
CD and PMD effects must be compensated to enable successful high-speed transmission over distance. For CD, compensation is currently realized by adding fixed dispersion-compensator modules to the system at installation, using highly trained personnel. The level of compensation must be matched to the given fiber link and accurate to within an acceptable dispersion tolerance that decreases dramatically with bit rate.
That presents an operational challenge for carriers as speeds increase. At 2.5 Gbits/sec, the dispersion tolerance is about 16,000 psec/nm, shrinking to 1,000 psec/nm at 10 Gbits/sec, and only 60 psec/nm at 40 Gbits/sec. Even at 10 Gbits/sec, this manual compensation process is labor-intensive, adding operational issues and cost.
At 40 Gbits/sec, the degree of compensation accuracy required demands precisely controlled automated compensation. At both 10 and 40 Gbits/sec, the eventual adoption of Multiprotocol Lambda Switching (MPλS) and all-optical crossconnects will force chromatic compensators to be dynamically adjustable.
Since the level of chromatic compensation must be matched to the length of the optical transmission path, the compensator will have to tune to accommodate changes as fiber links are reconfigured in the network. In this context, neither dispersion-compensating fiber nor manual tunable devices present a viable solution.
PMD places an economic burden on current 10-Gbit/sec long-haul systems that rely on expensive electro-optic regenerators to correct PMD signal degradation. In DWDM long-haul transmission, multiple optical-fiber impairments-including effects such as four-wave mixing, self-phase modulation, CD, and PMD as well as reduced SNR simply due to high-loss spans-act to degrade signal integrity.
While all of these mechanisms may impact the maximum permitted electro-optic regenerator spacing, the relative contributions of these impairments are dependent on the particular fiber transmission environment, and PMD dominates in a substantial percentage of the installed fiber plant. Average PMD values vary significantly across the existing installed fiber base depending on fiber type and age, with older fiber typically having much higher mean PMD values than the most recently deployed state-of-the-art fiber.
A significant amount of high-PMD fiber has been deployed to date-and in such environments, PMD often becomes a gating impairment, forcing the use of closely spaced regenerators to correct for PMD signal degradation. Regenerator equipment to perform optical-electrical-optical conversions places a serious economic burden on today's 10-Gbit/sec long-haul systems. Indeed, for very-high-PMD fiber, it becomes economically impractical to support 10-Gbit/sec operation given the number of very closely spaced regenerators required.
At 40 Gbits/sec, PMD poses a significant barrier to deployment in metro/regional environments and beyond. Metro/regional fiber links must be low-cost and therefore do not permit the use of electro-optic regenerators. Depending on the fiber type and age, PMD can result in large system penalties for relatively short fiber spans. For example, an 80-km link with average PMD of 0.5 psec/√km could have a power penalty of >5 dB resulting from first-order PMD alone.*
In this context, PMD compensation becomes essential to realize the performance required to support robust 40-Gbit/sec optical links. The metro/regional environment has more stringent economic constraints, limiting both capital outlay and operations resources for bringing up a link. Consequently, PMD compensators must be very inexpensive and easy to deploy-virtually plug-and-play solutions.
In the long-haul environment, 40 Gbits must be cost-competitive with long-haul systems at 10 Gbits. In the absence of PMD compensation, regenerator costs become prohibitive. Hence, PMD compensation is a critical technology to enable 40-Gbit/sec operation in metro/regional and long-haul networks.
New solutions will build active intelligence into integrated electro-optic interface products to provide adaptive compensation for these fiber-channel impairments. An emerging market area, OSP, applies DSP techniques commonly used in wireline applications to high-speed optical communications environments. The heart of OSP lies in the use of adaptive intelligence to optimize a complex set of design and performance interactions between electronics, optics, and the optical channel.
OSP provides active monitoring and dynamic compensation of the optical data stream, removing the need for external line conditioners and manual intervention at switch-on or when link characteristics change. This approach offers significant cost savings at 10 Gbits/sec and is essential for 40 Gbits/sec or higher, where fiber CD and PMD cannot otherwise be compensated in a pragmatic or economically viable manner. The development of OSP is analogous to the evolution of Ethernet rates, where DSP techniques that employ electronic equalization enabled Ethernet over unshielded twisted-pair copper to scale from 10 Mbits/sec to 1 Gbit/sec.
Signal processing techniques may be leveraged in several ways to compensate for CD. One approach applies electronic equalization, which is being developed in the context of PMD compensation, but also has applicability for correcting other impairments such as CD. An alternative is a tunable optical CD compensation device, with signal processing in electronics providing active feedback control that enables automatic setup and dynamic optimization of an optical link.
The quality of the received data is measured, and this measurement is used in a microprocessor-controlled feedback system that tunes an optical chromatic dispersion compensation module (DCM). Figure 5 shows an optical DCM adjusted via a feedback control processor, using an error signal generated within the receiver chain. Such a solution demands a closed-loop system approach, beginning with optical system designers defining the high-level control-loop architecture requirements. Furthermore, the data-monitoring function, control algorithm development, and interface with the optical DCM require the tight coupling of optics, electronics, and signal processing capabilities, representing the essence of OSP. Figure 2b shows the signal of Figure 2a following transmission through 80 km of SMF (~17 psec/nm/km) after being corrected with a tunable optical dispersion compensation device adjusted to optimize the eye opening.
The closed-loop system approach provides a simple plug-and-play solution to the dispersion compensation problem and is particularly advantageous in a metro/regional environment, where limited staff with fiber transmission expertise is available for system installation. The same DCM could be used to bring up a new fiber link, regardless of distance (e.g., with a 50- or 80-km link length), without requiring any user intervention to tune.
Rather, the DCM would be set automatically via electronic feedback control that optimizes the quality of the received optical signal for the given link. In future systems, with MPlS capabilities and optical crossconnects, a tunable optical DCM device with electronic feedback control would be well suited to dynamically compensate for significantly changing path lengths on optical links.
Furthermore, in traditional long-haul environments, electronic feedback control of tunable optical DCM devices may be used to address residual dispersion and dispersion slope mismatch. Because fixed DCMs are only available in discrete values (typically 170 psec/nm), residual dispersion between ±85 psec may result, leading to a system penalty that becomes significant at 40 Gbits/sec.
In DWDM systems, channels at the edge of the wavelength band may incur additional penalty due to dispersion slope mismatch across the band. Applying feedback-controlled tunable optical DCM devices per wavelength resolves both these challenges. The alternative, performed purely in the electronic domain, offers potential to be very effective in some applications, providing a significantly lower-cost solution.
Although adaptive PMD compensation to date has been approached primarily in the optical domain, the adoption of these optical devices has been slow due to their high cost and large size. Electronic IC solutions are now being pursued to reduce both cost and footprint. One logical implementation is based on a tapped-delay-line architecture (see Figure 6). In this electronic equalizer, the various tap weights must be dynamically adjusted in response to changing PMD statistics of the fiber link, calling for advanced DSP control algorithms. Optical system simulation is required to determine overall design requirements for such a compensator system, and implementation at 10 and 40 Gbits/sec requires high-speed broadband electronic design expertise. Again, an OSP solution is required, with optical, electronic, and signal processing technologies working hand-in-hand.
Figure 4b shows the signal from Figure 4a after being compensated by an electronic tapped-delay-line equalizer, which mitigates both DGD and higher-order PMD. To compensate more severely impaired signals, the addition of a decision feedback equalizer may be required.
The electronic equalizer not only compensates PMD, but also provides correction for other intersymbol interference-based signal distortion mechanisms such as SPM, transmitter patterning, and even CD. In the context of CD, the electronic equalizer is expected to have a more limited compensation range than optical devices but would be well suited as a low-cost solution for residual dispersion and slope mismatch.
Fixed optical line conditioners such as dispersion compensation fiber (DCF) suffer from slope mismatch as well as residual dispersion, making them insufficient for the precise compensation required at 40 Gbits/sec-and introducing significant system penalty at 10 Gbits/sec. Multichannel tunable optical compensators such as fiber Bragg grating and etalon-based devices may be set to eliminate residual dispersion at the center of the band but remain subject to a slope mismatch across the band that results in system penalty for outlying wavelength channels.
The use of electronic compensators based on electronic equalization, in conjunction with optical compensator devices, significantly reduces system penalty due to slope mismatch and/or residual dispersion. Low-cost electronic equalization enables the level of precise dispersion correction required for 40-Gbit/sec operation and reclaims substantial system margins, yielding reduced capital costs for both 10- and 40-Gbit/sec systems.
For scaling optical networks to higher transmission speeds, OSP brings substantial benefits in terms of reducing operational complexity and capital cost. Only a fully automated plug-and-play solution is viable for 40-Gbit/sec deployment in metro/ regional environments, where carrier costs must be a fraction of those associated with long-haul links. By harnessing the power of signal processing, OSP makes CD and PMD compensation transparent to the metro system provider, permitting the current mode of deployment for 2.5-Gbit/sec systems to be extended to 40 Gbits/sec.
In the long-haul environment, OSP technology provides significant cost savings at 10 Gbits/sec and is a key enabler for viable 40-Gbit/sec operations. A traditional approach involving manually installed fixed-line conditioners to correct CD sacrifices system margin at 10 Gbits/sec and simply will not work at 40 Gbits/sec.
OSP technology enables the development of intelligent optical subsystems that remove the operational issues associated with compensating CD at high bit rates. Furthermore, application of adaptive electronic equalization techniques enables low-cost, small-footprint PMD compensators, dramatically reducing regenerator-based capital outlay and mitigating other signal distortion mechanisms.
All these advances result in powerful gains in system performance and resilience. Auto-adaptive compensation techniques further ensure the system is consistently operating at peak performance over its lifetime, making available the system power budget that normally is lost to "system margins."
Dr. Laura Adams is the chief technical officer for the optics division, Jonathan King is manager of optical systems, and Sudeep Bhoja is the senior design engineer at Big Bear Networks Inc. (Milpitas, CA). They can be reached via the company's Website, www.bigbearnetworks.com.
- C. Poole and J. Nagel, "Polarization Effects in Lightwave Systems" (Eq 6.17), Optical Fiber Telecommunications IIIA, edited by I. Kaminov and T. Koch, Academic Press, 1997.