Although commercial components are just emerging and standards have yet to be set, CDWM offers distinct advantages for short-haul, unamplified networks. Lower equipment costs stem from the use of uncooled lasers and other components manufactured to less stringent tolerances than required for DWDM.
chief technical officer
The promise of metro/access networks and their ability to flourish as a huge market for fiberoptics is a popular topic for discussion. Numerous experts have identified the high cost of components, and specifically dense wavelength-division multiplexing (DWDM) components, as the primary reason for delay in the rapid growth of this market. However, a metro/access network based on coarse WDM (CWDM) can provide the balance of price and performance needed for this unique market. CWDM enables metro/ access in much the same way that DWDM enabled the long-haul market.
Both DWDM and CWDM are types of WDM: DWDM is the implementation of WDM over long distances and CDWM is the implementation of WDM in metropolitan and access markets. The different requirements of these two markets frame the various architectures and drive the performance requirements of the system components (see Fig. 1).
DWDM for long-haul
The development of the erbium-doped fiber amplifier (EDFA) has been the primary enabler for the proliferation of high-bandwidth long-distance data networking by significantly reducing the need for costly traditional 3R (reamplification, reshaping, retiming) regeneration equipment. The EDFA`s inherent ability to simultaneously amplify multiple signals independent of the wavelength and bit rate allows network operators to offer low cost capacity in DWDM systems.
The architecture of long-haul DWDM systems demands high performance components. The market trend is clearly towards deployment of all-optical DWDM systems with more channels, longer spans, and wider wavelength spectrums, forcing component manufacturers to deliver yet higher performance and more expensive components.
The components in the long-haul infrastructure that are the primary cost drivers of the network include regenerators, optical amplifiers, thermally-controlled distributed-feedback lasers, multiplexers, demultiplexers, optical add/drop multiplexers, switches, cross-connects, gain equalizers, dispersion compensators and receivers. Despite the high costs incurred in building, operating, and maintaining a long-haul network, the network delivers cost-effective capacity by virtue of the large number of customers it serves.
Metro/access equipment costs
Although the demand for bandwidth services is apparent in the emerging metro/ access market, DWDM has not been widely deployed in this market. The metro/ access market does not require the same bandwidth and distance requirements as the long-haul network.
The challenge for metro/access networks is in distributing the capacity delivered by the long-haul network core, and aggregating from the network edge back to the long-haul core. Multiple topologies have been proposed by system vendors to provide solutions to this challenge. Although the problem is described from various perspectives and presented with numerous innovative solutions, there is no single correct solution. There is, however, a common thread among all of these solutions: the cost per bit is dominated by equipment costs.
To strike a balance between price and performance in metro/access applications, the costs of the components must be reduced, while providing adequate bandwidth. CWDM effectively takes what is great about WDM (multiple channels of light on a single fiber) and effectively implements it with lasers that are significantly less expensive than their DWDM counterparts.
Laser diodes used for WDM systems (CWDM or DWDM) are predominantly distributed-feedback (DFB) chips. This structure and the materials used to make it cause the output wavelength of the chip to change as the temperature changes. In practice, the typical wavelength change of a DFB chip is 0.08 nm/°C. In DWDM systems, it is necessary to use costly packaging techniques (butterfly housings with thermo-electric coolers) to prevent the wavelength from drifting. In CWDM systems, a significant cost reduction of the system is achieved by using non-thermally controlled (uncooled) lasers (see Fig. 2).
Cost savings are a direct reflection of the packaging differences between DWDM and CWDM lasers. The difference between the packaging of DWDM lasers and CWDM lasers is analogous to comparing a Formula One race car and a standard passenger car. Each Formula One laser is assembled with the utmost care to optimize the performance. Alternatively, CWDM lasers can be produced with a much higher yield and at a cost appropriate for the metro/ access. Specifically, TO can and mini-dil type packages used for CWDM lasers are routinely manufactured in automated facilities.
Higher tolerances, lower prices
Coarse WDM systems are tailored for applications over modest distances that do not require an EDFA. Once the constraint of the EDFA bandwidth has been removed, wavelengths can be distributed over a wide region and spaced far enough apart to prevent the signals from drifting into each other. Over a typical operating temperature of -100°C to +700°C, the wavelength could wander slightly more than 6 nm.
In addition, the laser manufacturer requires a wafer-to-wafer tolerance of ±3 nm to produce the laser chips inexpensively. Similarly, the WDM filters required to separate and combine the signals have various tolerances. Once all the tolerances are added, CWDM signals should be spaced approximately 20-nm apart to ensure the maximum usable bandwidth while keeping the signals from interfering with each other.
To become widely adopted, CWDM systems will need low-cost multiplexers, demultiplexers, add/drops, and switches. These components cannot simply be retrofitted from DWDM systems. Particularly needed are multiplexers and demultiplexers with low loss, high isolation, and proper channel spacing (see Fig. 3). Loss in excess of several dB through the mulitplexer and demultiplexer would drastically reduce the distance a signal could travel. Proper channel spacing and high isolation between the channels is necessary for reliable service. Clever companies are finding ways to match the cost reductions associated with uncooled lasers to cost reductions in all the other components required for CWDM systems.
With some frequency, CWDM proponents are presented with the following hypothetical question: "If DWDM filter prices fall, doesn`t that eliminate the need for CWDM?" While there will always be applications for DWDM, it`s unclear that price erosion in DWDM will have a dramatic effect on the need for CWDM.
The reason is straightforward. Unless a new structure and material is created to replace the DFB laser in DWDM systems-which seems unlikely in the next several years-thermally controlling these DFB lasers will still be required for DWDM systems. Given the high volume production scenario of both cooled and uncooled lasers, it is not likely that DWDM lasers can match the packaging simplicity and inherent cost differential with CWDM lasers.
It is conceivable that CWDM lasers will ultimately be produced in the same packages and factories as lasers made for DVD players (the packaging similarities are striking). DVD lasers are manufactured in the millions for very low costs. This long-term possibility of continuing improvements in the price/performance balance makes CWDM an ideal fit for emerging metro/access applications.
Spacing and benefits
While metro/access systems do not require as much bandwidth as long-haul systems, a significant amount of bandwidth is still needed. So the question stands: How much bandwidth can a CWDM network support, given that the lasers are spaced so far apart?
A number of vendors are proposing to use the entire 1300- to 1600-nm transmission window of optical fiber to support CWDM. Of course the reduction in loss and increase in dispersion going from 1300 to 1600 nm must be considered, both with respect to installed fiber and new fiber proposed for metro/access applications. In a channel plan that spaces lasers 20 nm apart and extends from around 1300 to 1600 nm, a system can accommodate more than 10 wavelengths. At 10 Gbits/s per channel, CWDM networks should cost-effectively deploy at least 100 Gbit/s links in a metro/access application.
At present, CWDM is in its infancy. Uncooled lasers operating at 2.5 Gbit/s and spaced 20 nm apart are available from a number of vendors. Four-channel mux/demux modules are available. Under development by several manufacturers are 10 Gbit/s uncooled lasers, but few are commercially available. Multiplexers and demultiplexers with more than four channels are under development, but again are not largely available. Furthermore, switches and add/drops for CWDM, which will be required, have not yet emerged.
CWDM systems enable metro/access in much the same way that DWDM systems enable the long haul. Additional work is needed to make the right components available and to set the standards, but it is clear that CWDM systems offer are a potential solution to some of the challenges of the metro/access market.
James Campbell is the chief technical officer of Tsunami Optics, 935-F Sierra Vista, Mountain View, CA 94043. He can be reached at 650-940-6800 or email@example.com.
FIGURE 1. The goal of a DWDM system is to maximize distance without electrical regeneration, while distributing the cost of an amplifier across the maximum number of wavelengths (top). The goal of a CWDM system is to minimize the cost of components in systems where the distance is modest and EDFAs are not required (bottom).
FIGURE 2. In DWDM systems, the wavelength channels are within the bandwidth of an EDFA (1530 to 1625 nm). More expensive, cooled lasers must be used to prevent the wavelengths from drifting outside this window or from interfering with each other (top). In CWDM systems the channels can extended over the entire 1280- to 1625-nm band using much less expensive, uncooled lasers (bottom).
Setting CWDM standards
It is important to establish standards for CWDM because those standards will support the rapidly accelerating metro/ access market. Defacto standards for metro/access networks are emerging within the market already: IEEE committees are setting standards for 10 Gbit/s Ethernet and beyond.
Can logic come from standards chaos in the metro/access market? Only if the participants (components suppliers, systems builders, and service suppliers) get together and agree on self-regulation.
Coarse WDM muxes/demuxes and lasers are already moving toward a self-imposed standard for purely pragmatic reasons. Systems architectures are best served by standard components and test equipment. Wavelengths are being selected by manufacturers to provide "just the right amount of bandwidth" so that uncooled lasers can operate over a wide temperature range and still stay within the CWDM channel width.
Currently, the channel width is specified at 20 nm nominal on centers of 1491, 1511, 1531, etc. to 1611 nm. In the 1300 region, IEEE Ethernet is defining channel width as 20 nm, but on wavelengths of 1290, 1310, 1330, and 1350 nm. What happens when they meet in the 1400-nm region?
Standards for metro/access must be established. To this end, Tsunami Optics (Mountain View, CA) is forming a CWDM users group to begin addressing metro/access standards chaos. For more information, contact firstname.lastname@example.org.