Multiplexers bring DWDM to metro/access markets

Jerry R. Bautista

The need for increased bandwidth has driven system providers to offer ever-increasing channel counts at narrower channel spacings. Long-haul systems continue to be deployed with at least 16 channels and often with the capability of 40 or more channels. Reliability and compatibility with optical amplifiers are essential.

This need for optical-amplifier compatibility, in fact, has required narrower channel spacings with increasing channel count to remain within the current optical amplifier window of 1530 to 1565 nm. There are programs to extend this window to roughly 1610 nm and also efforts to develop new amplifiers with even broader windows such as Raman amplifiers (see related article on p. 27). Given the existing transmission window, channel spacings of 200 GHz (1.6 nm) allow for 16 channels, while spacings of 50 GHz allow for 80 channels. Systems deployed roughly two years ago had channels on 200-GHz centers, while systems to be deployed next year are currently under development with 50-GHz channel spacings.

MARKET DRIVERS

From a dense wavelength-division-multiplexing (DWDM) system perspective, metro/access environments are more difficult to generalize because the challenges and topology vary broadly, as do the approaches and number of system providers. Although the metro/access market is relatively small in size now, it should grow to more than $2 billion annually by 2005, according to a study by Frost and Sullivan. Ease of provisioning, first-installed cost, and the ability to grow bandwidth-or channel count-in a scalable fashion are critical in this environment. Although channel counts tend to be 16 and below, channel spacings varying from 200 to 50 GHz will soon be available.

Recent market data indicate several trends. First, channel counts are indeed increasing. Further, there is a migration from 16 and below to 32 channels and above, predominantly in the long-haul market. However, there is not likely to be a significant reduction in the 16-and-below channel-count market because of the entrance and increased deployment of DWDM systems in the metro/access market. Finally, the obvious trend of increasing overall system revenues year upon year is clear.

TECHNOLOGY COMPARISON

The challenges to a DWDM component supplier in the metro/access market are to provide reliable, high-quality components that allow for flexibility of channel plan and spacing at a good value to the customer. Future upgrade paths both in spacing and channel count tend to drive the selection of a technology platform rather than one of a specific DWDM filter to provide a targeted mux/demux function.

Several basic technologies are available to produce suitable DWDM mux/demux filters. Venture capital funds are increasingly available to the optical component startups, driving more varied approaches and further product distinctions. At this point, there does not seem to be a dominant DWDM mux/demux technology that provides the best of all customer requirements in terms of performance, cost, reliability, and flexibility.

Primary performance points of comparison are insertion loss (the efficiency with which individual wavelengths are separated), channel isolation (the ability to accurately select individual wavelengths), environmental stability (spectral stability in the face of temperature variations or vibration, for example), and pass bandwidth (the width of the window for any given channel). Polarization-dependent effects as well as channel-to-channel insertion loss uniformity are also often parameters of interest. Finally, compliance with Bellcore GR 1209 or 1221 reliability guidelines is often specified.

There are three basic approaches: thin-film dielectric devices, planar waveguides, and fiber-based components (gratings or interferometric fiber devices such as Mach-Zehnder elements). Each has strengths in the filter technology market (see Fig. 1).

Thin-film dielectric devices are probably the most broadly deployed filters for DWDM systems with channel spacing from 400 to 200 GHz. It is a very mature technology offering good temperature stability, channel-to-channel isolation, and a broad passband. To provide such performance, 200 or more layers of material are deposited in a carefully controlled manner on a glass substrate in large deposition chambers. Chips are diced, polished, and precision-mounted in metallic housings along with collimators to yield a wavelength-specific free-space device. The primary challenge for this technology is to decrease channel spacing to 100 GHz and lower, as well as increasing channel count economically beyond eight.

Planar waveguides are composed of a few glass layers of carefully controlled composition upon a silica or silicon substrate. These layers are patterned and etched using variants of standard semiconductor process techniques: photolithography and reactive ion etching. Several devices can be produced on a single wafer. The primary challenge is to control layer thickness, composition, and defect inclusions for devices that are typically much larger than those produced in standard IC facilities. These devices lend themselves to high integration and consequently large channel counts. Channel spacings are typically 200 to 100 GHz, although 50-GHz devices are under development. Temperature stability is often an issue, requiring actively heating the devices above ambient. Insertion loss is often compromised to some degree as well.

Fiber-based devices are typified by long or short-period gratings or interferometric structures such as Mach-Zehnder configurations. These devices perform better, particularly the latter, at narrower channel spacings and moderate channel counts (fewer than 16). Insertion loss and uniformity is very good for these components as they are fabricated from standard single-mode fiber. Channel spacings as narrow as 2.5 GHz (0.04 nm) have been demonstrated using this technology.

CHALLENGES AND LIMITATIONS

The selection criteria for current metro/access applications often require a hybrid approach because of limitations of a particular technology platform or the need to provide a graceful path for bandwidth expansion. Few DWDM mux/demux filter suppliers are able to provide such a technology hybrid, and customers must perform the integration themselves. Alternatively, this opportunity sets the stage for a situation where competitors must cooperatively supply a need for a given customer and perform the integration jointly.

Another major consideration is product reliability. Increasing channel counts and narrower channel spacings lead to higher bandwidths, which require careful consideration of failure rates and overall product reliability to avoid extremely costly service interruptions. At present two major standards set reliability guidelines for optical components: Bellcore GR1209 and GR1221.

Perhaps the most critical individual test is the high-temperature/high-humidity storage test. The device must be able to withstand 85°C and 85% relative humidity with less than 0.5 dB insertion-loss variation at the end of 500 hours and provide information only at 2000 hours. The suggested sampling plan is for zero failures in 11 samples, one failure in 18, and two in 25. Other tests involve storage and temperature cycling as well as mechanical tests such as impact and vibration.

In addition to specifications, customer negotiations often include allowable variations in optical properties for various reliability tests. System considerations provide allowable variation and failure rates for each major element in the assembly.

INTERLEAVING TECHNOLOGIES

An increasingly common solution to achieving narrow channel spacings (50 GHz and narrower) is the use of "interleaver" technologies. Basically, two separate mux or demux devices with twice the channel target spacing are combined to cover the entire operating window by interleaving them-one mux or demux covers the odd channels while the other covers the even ones (see Fig. 2).

This approach allows a technology that performs better at a wider channel spacing to address a narrower one. In other cases, long design cycles to fabricate components that do perform well at the narrower channel spacing, as well as another round of Bellcore testing, can be avoided. The cost of the resulting assembly may then be lower as components with wider channel spacing are more mature, typically having higher yields and consequently lower cost. This interleaving technology can take one more step with a 1 × 4 front-end device (see Fig. 3).

Such a device allows channels to be easily added in banks of four. This scalable approach to increasing bandwidth is of particular interest to the metro/access market, allowing a "pay as you grow" approach and the potential for increased flexibility in optical network provisioning.

Whether using a 1 × 2 or 1 × 4 interleaver, fiber-based devices provide a balance of cost, reliability, and high performance. These devices allow for extremely low insertion loss with very good isolation and bandwidth flexibility. Furthermore, very narrow channel spacings can be achieved with an all-fiber, Mach-Zehnder device (see "Fewer channels, but they cost less," p. 12).

FIGURE 3. A 1 × 4 front-end device allows channels to be easily added in banks of four and may be of particular interest to the metro/access market because of the flexibility it provides to incrementally expand network capabilities.

Fewer channels but they cost less

While the emphasis today is upon very narrow channel spacings for DWDM systems, not all applications may require the associated cost and complexity of network management. A number of system providers are exploring the development of narrow wavelength-division multiplexing (NWDM) systems in which four channels are spaced as far as 10 nm apart in the1530-1560-nm window (see Fig. 1).

In this manner, less-expensive lasers with relatively wide transmission bands and reduced center wavelength precision can be employed. Similarly technologies such as narrowband fused fiber couplers can be used as the building blocks for mux and demux components. These devices are entirely passive, small, and have very low insertion loss (see Fig. 2). They offer good performance and are consistent with the emphasis on cost and simplicity.

FIGURE 2. Two 10-nm narrowband demultiplexers could follow a 5-nm demultiplexer in a NWDM system.