‘Light WDM’ stretches metro optical capacity

Rapid and continuing growth of broadband-based services is completely transforming the telecom environment. To provide a perspective, the market-research firm Infonetics Research wrote in its Broadband CPE Report (June 2006) that it expects 245 million DSL subscribers worldwide by 2009, with 70 million cable broadband subscribers, and more than 53 million IPTV subscribers. This is from an almost zero base at the start of the new millennium, and is equivalent to creating a market the size of the population of the European Union in a decade. In North America alone, Infonetics Research expects the number of IPTV subscribers to increase more than 40-fold between 2005 and 2009.

Such growth is increasingly driven by service convergence and bundling, as carriers and service providers begin to migrate to next-generation IP/MPLS/Ethernet networks to support a unified service/application environment encompassing voice, video, and data services over converged fixed/mobile networks.

An inevitable consequence is a huge increase in the bandwidth requirements placed on carrier networks, but particularly on the peripheral and customer-facing access and metro networks. So premium broadband services such as IPTV-clearly identified by carriers as an imminent major commercial opportunity-will require several megabits/sec per channel per subscriber. While 8-Mbit/sec DSL is already becoming common, the widespread introduction of HDTV and faster access technologies such as VDSL2 and FTTH over the next few years will make 20 to 30 Mbits/sec commonplace.

Metro access networks will take the brunt of this demand, as they provide the collector and distribution rings for the access networks and also backhauling to PoPs and long-distance networks for a wide variety of service providers and applications-corporate networking, storage networking, and wireless backhaul, for example. Unfortunately, in many locations, these networks are not rich in fiber, and pulling new fiber is too expensive or difficult for many carriers to contemplate.

To overcome this potential metro bottleneck, many carriers have turned to coarse wavelength-division multiplexing (CWDM). This technology has a lot of attractive attributes for the metro access area. It is very inexpensive compared to the DWDM originally developed for core and long-distance networks, as CWDM has to conform to much less demanding specifications than DWDM-for example, channel spacing is much wider, which means that laser wavelength temperature drift is less critical, and lasers can be uncooled and cheaper.

CWDM generally uses standard-based compact small-form-factor pluggable (SFP) optics, and eight wavelengths per fiber are easily supported, giving a significant increase in bandwidth capacity. The equipment can be made very compact, with low power consumption-desirable characteristics in peripheral networks such as the metro access. As metro networks are numerous, volumes are potentially high, and thus long-term cost trends are very favorable.

Metro CWDM transport platforms are thus becoming an important part of the optical network market. However, CWDM does have some limitations as a long-term solution for metro applications as bandwidth demand continues to increase. The most obvious is that it doesn’t support very many wavelength channels. The ITU-T Recommendation G.694.2 Standard Grid defines 18 discrete channels, 20 nm apart and centered from 1,270 to 1,610 nm. Unfortunately, two of these, at the edge of the transmission window, can suffer from Rayleigh scattering and so tend not to be used, bringing the usual total down to 16.

Even this overstates the case, however, as these wavelength channels span a region around 1,380 nm where much of the installed base of metro fiber suffers from high attenuation due to the so-called water-absorption peak at that wavelength. Currently, most vendors use only the upper eight or nine wavelength channels above this peak. Modern zero- or low-water-peak fibers do not have this problem, of course, but are rare in metro networks, and the cost of replacing old fiber would negate the point of using CWDM in the first place. Furthermore, as a result of such limitations, the additional 10 channels will lower the overall performance of the system including the original eight.

So, for practical purposes, eight wavelengths are probably as far as CWDM will go for most carriers. A further constraint on system capacity is that CWDM SFP optical transceivers cannot currently support data rates in excess of 2.5 Gbits/sec, although this may improve in the future. Finally, CWDM was specifically designed not to use optical amplification (to keep down the cost), and so spans are much shorter than for DWDM, typically 50 to 80 km. Although the shorter span length is not too serious a problem for many carriers, there are situations where longer spans would be needed, and the inability to meet such requirements is an obstacle to using CWDM in a metro regional role by, say, smaller carriers with lower bandwidth needs.

One could advocate using DWDM as a complete high-capacity metro approach; this would provide many more wavelengths than CWDM and associated optical transmission rates are much higher-10-Gbit/sec interfaces are widely available, for example, and optical amplification could be used to obtain longer spans. However, cost remains an issue as, very roughly, a DWDM platform costs anywhere from two to five times a CWDM platform on a per-wavelength basis.

Cost efficiency is another issue. A full DWDM deployment would be overkill in capacity terms in many metro access networks and would result in high fixed capex and opex costs due to the large, expensive common parts and the need to house, power, and cool such equipment.

It would be wrong to view CWDM and DWDM as opposing metro approaches. Each can be appropriate for different metro applications, depending on the circumstances. It is more useful to see if it is possible to combine them into a hybrid platform to obtain the benefits of both in a widely applicable solution.

Such an approach should be viable, provided that one starts from an appropriate low-cost CWDM platform and adds a suitable low-end DWDM capability, rather than trying to engineer down an existing, more costly DWDM platform. The resulting C/DWDM platform could be referred to as providing “light WDM” (LWDM).

Looking at such a combination in more detail, the CWDM platform should be designed from the start to allow development to a hybrid offering, as already mentioned. The CWDM system’s capabilities then could be expanded with the following: