Coarse wavelength-division multiplexing (CWDM) technology promises to contribute significantly to the growth of optical networking by providing a low-cost alternative to DWDM needed for mass deployment of optical metropolitan networks. With support of several key emerging technologies, the metro-core and access networks are poised to play a major role in delivering new and improved telecommunication and data-communication services to customers.
Fundamentals of CWDM
Systems based on CWDM are typically designed to have widely separated wavelengths stretching from 1300 to 1620 nm. CWDM systems resemble their high-density cousin, DWDM, in that they have sources operating at different wavelengths, a filter to combine the wavelengths onto a single fiber, and another filter to separate the wavelengths at the receiving end. Also, adding and dropping of wavelengths in the middle of the system can be accomplished with the same technology used in DWDM systems. The main difference is that in DWDM, the channel spacing can be as close as 0.2 nm (25 GHz), while the channel spacing for CWDM is typically 20 nm.
The first component necessary to realize low-cost WDM transmission is the uncooled coaxial DFB laser, which costs only a third of a cooled DFB butterfly-packaged laser typically used in DWDM systems (see Fig. 1). The inherent flexibility of wide wavelength spacing in CWDM allows the use of uncooled lasers and therefore drives the cost down significantly.
This cost savings stems from the complex packaging required in DWDM systems to stabilize the wavelength sufficiently to keep neighboring channels from interfering with each other. A DWDM laser is typically a DFB, which is mounted in a 14-pin butterfly package. Optical reflections and temperature changes in the laser, as well as changes in drive current, cause the wavelength of the laser to fluctuate. Thermoelectric coolers, built-in optical isolators, and photodetectors are all necessary components to operate the DFB at a stable wavelength. In addition, cost-adding external wavelength lockers are sometimes required. Finally, lack of automation for DWDM laser assembly increases the manufacturing cost.
Uncooled coaxial lasers used in CWDM, on the other hand, are allowed to operate over a wide temperature band, ranging from 0°C to 70°C (see Fig. 2). Currently, manufacturing tolerances at room temperature are typically ±3 nm for standard devices and ±2 nm for premium devices. Temperature dependence of the center wavelength, at 2 mW of fiber output, is typically 0.1 nm/°C, which gives a total wavelength drift of 7 nm.
Data rates are now up to 2.5 Gbit/s, with the 10 Gbit/s speed on the horizon. Standard wavelengths available today range from 1471 to 1611 nm, with the prospect of extending the range down into the 1300-nm region.
Another key performance parameter of the uncooled lasers is dispersion penalty (see Fig. 3 and Table 1). Laser manufacturers typically achieve a distance of 60 km over standard single-mode fiber at 2.5 Gbit/s with dispersion coefficients varying from 13 to 19 ps/nm*km depending on the transmission wavelength. Efforts are under way to optimize the DFB chip design to extend the distance limitation resulting from the dispersion penalty.
In addition, the passive filter used for wavelength multiplexing is another critical CWDM low-cost component. Thin-film filter technology is currently the technology of choice for low-cost, high-performance CWDM systems. Using multilayer glass-substrate technology, filters can typically achieve optical isolation of 30 dB with minimal insertion loss. A channel passband greater than 11 nm is necessary to allow for laser manufacturing tolerance and drift caused by peak wavelength temperature dependence.
One concern is the cost of thin-film filters as a function of the filter`s figure of merit (FOM). The FOM is the ratio of passband at an upper location around the peak: from 0.5-dB bandwidth to approximately 20-dB. For CWDM, the FOM is calculated to be around 0.65, with 13-nm passband and 20-nm spacing. At this FOM, the CWDM filter realizes an approximate 20% cost savings compared with DWDM channels. In the future, additional cost savings can be achieved by increasing the wavelength spacing beyond the standard 20 nm. This relaxed wavelength tolerance means that many of the stringent optical specifications that DWDM filter manufacturers routinely wrestle with, such as insertion-loss ripple, polarization-dependent loss, or temperature coefficient of center wavelength, are no longer major obstacles for CWDM filters.
Integrated solutions are also showing up in the market. Uncooled OC-48 DFB CWDM transmitters and transceivers are becoming available for metro-WDM-network applications. Developed for a target distance of less than 60 km over conventional single-mode fiber, the modules are designed specifically for 13-nm-wavelength windows and operate over a wide temperature range. These transmitters and transceivers are packaged in the popular SONET form factors, including traditional 2 × 10, 1 × 9, and 2 × 9 packages, as well as the newer small form factor (SFF) transceiver packages. These modules feature an integrated laser driver and monitors, and allow designers to interface with popular framer chipsets via PECL outputs and inputs.
Which metro and access spaces?
A metropolitan area network (MAN) is a term broadly applied to the interconnection of networks within a city into a single larger network. It is often structured as the interconnection of several local area networks by bridging them with backbone lines. Today, however, with the explosion of the Internet, intranets, and extranets, the major segments of the metro network are often split and configured to be the metro core, the metro access, and the enterprise segments.
Historically, the metro network as a whole is often classified as the local-exchange network that interconnects the carrier`s central offices or points of presence. It is clearly differentiated from the interexchange (long-haul) network that interconnects cities and major traffic hubs, or the local access network that reaches out from a carrier`s central office into individual homes and businesses (see Fig. 4).
The distinction made between the metro-core and metro-access market segments is not framed on the traditional spatial architecture, where one carrier with superior equipment infrastructure and market presence can bury the newcomer. Instead, it follows the competitive model that is based on the flow of bandwidth being either distributed or backhauled. The distinction in traffic patterns is such that, while the metro-core network sustains more evenly distributed, meshed traffic-like structures connecting central offices, ISP hubs, and other "headends," the metro-access network brings backhauled aggregated traffic up to the metro-core nodes.
CWDM is an alternative to expensive and complex DWDM-based architectures because it provides the opportunity to continue the momentum created by DWDM technologies toward an all-optical network. Metro-DWDM systems are often seen as being inappropriately engineered for the metro space; they were originally designed to meet the demands of the long-distance service providers. The advantage of DWDM eliminating costly regenerators in long-haul networks does not apply to metropolitan networks, where optical amplifiers are either not required or low-cost gain block modules with inexpensive uncooled laser-pump modules easily meet the distance requirement in most metro rings.
While the low cost of laying fiber may be an attractive DWDM alternative in certain metro areas, it has proven too expensive in other segments. Fiber installation remains too costly in many places for a variety of reasons, including urban congestion, permits, codes, and the like. In addition, next-generation DWDM-based systems incorporating a thin layer of SONET multiplexing to provide narrowband services and next-generation SONET systems like multiservice provisioning platforms (MSPP) are being advocated by several vendors as having the potential to provide meaningful solutions to the metro market when properly adapted.Capacity upgrades using standard SONET add/drop multiplexers (ADMs) are limited to a certain bandwidth and become too expensive to achieve OC-192 data rates. The result: next-generation ADM/DWDM networks are still too complex and need to see a significant price reduction before they could meet most service provider requirements for the metro space.
Although low-cost metro EDFA modules are not quite ready to be deployed or, in many cases, not needed in most short-haul applications, low-cost Gigabit Ethernet transceivers promise to contribute significantly by extending LAN boundaries into the metropolitan area through 4, 8, or 16 CWDM channels. CWDM will support those high-end LAN internetworking services along with such channel interfaces as FDDI, FICON, and ESCON.
The growing interest in using Ethernet in the MAN stems from its easy interworking with traditional LAN technologies. The Ethernet framing standard is well-known, popular, and capable of helping to defray the cost of maintaining complex hybrid SONET/DWDM next-generation networks. The cost associated with training personnel to manage optical network technologies can be very high. The recurring expenditures involved with operations, administration, maintenance, and provisioning will be greatly reduced by using the familiar LAN-based architecture.
An attractive proposition is to use CWDM and 10-Gigabit Ethernet transceivers to develop rate-adaptive CWDM systems. Here the data-signal rate could be adapted to match the fiber`s link-loss budget by optimizing the rate for a given insertion loss. Hence, the new element introduced in the WDM equation is the fact that system developers are starting to acknowledge that Ethernet can indeed be pumped up to terabit speeds, and CWDM technology is well positioned to take advantage of this opportunity.
Other emerging technologies, such as the recently introduced low-water-peak fiber will help achieve similar results in different segments of the metropolitan optical network. This new fiber, when used with CWDM, offers real opportunities for revenue enhancements and cost reduction in metro core and access networks. Low-water-peak fiber opens the 1400-nm band and makes whole optical fiber transmission spectrum from 1300 to 1600 nm available.
Using this extended CWDM architecture, system solutions could be easily implemented to scale from 4 to 16 channels to deal effectively with the fiber-exhaust situation in certain areas, and help alleviate cost-per-bit and cost-per-connection problems. Using CWDM, a mix of services and diverse modulation formats can be easily provided to the end user on a single fiber without interference.
The author extends his gratitude to the ExceLight applications engineering staff for their contribution. Acknowledgment is also due Martin Mastenbrook of Quantum Bridge Communications for his technical guidance.
Jean-Jacques (JJ) Petiote is a senior applications engineer at ExceLight Communications, 4021 Stirrup Creek Drive, Durham, NC 27703; www.excelight.com. He can be reached at 919-361-1633 or email@example.com
FIGURE 1. The pigtailed coaxial DFB laser diode package contains a fiber pigtail attached to an optical isolator and an aspherical lens, which focuses the light onto the input isolator. The four pins are connected to the LD anode, the LD cathode, the PD anode, and the PD cathode.
FIGURE 2. The solid line indicates the wafer manufacturing tolerance, which may vary to up to plus or minus 3 nm. The dotted line represents the total wavelength window due to both wafer variations and temperature effect.
FIGURE 3. The dispersion penalty generally limits the transmission distance for DFB laser diodes over standard SMF to approximately 60 km at 2.5 Gbit/s.
FIGURE 4. Metro network is a term loosely applied to the interconnection of several types of local network segments.