By Jim Sauer
In a typical long–haul DWDM network, static OADM architectures make sense. For metro, a flexible architecture that combines band and per–channel filters often provides the most dynamic solution.
As larger numbers of DWDM channels are deployed, practical use of fiber bandwidth becomes essential. The ability to selectively add and drop individual channels or groups of channels in a complex traffic pattern without regeneration represents significant cost savings. With the proper node and network design, optical add/drop multiplexers (OADMs) provide this flexibility and efficient traffic management, significantly reducing cost. Carriers must consider the type of OADM architecture as well as the specific requirements of the network to design an optimal solution.
The optical filter is a basic building block of DWDM systems, providing the necessary wavelength selection to isolate independent channels. The transmission characteristics of the optical filter have a significant effect on the performance of the system as a whole. For example, in DWDM systems the filter passband characteristics and its roll–off slope must be considered in relation to the wavelength separation to minimize crosstalk. In addition, for a system using more than one optical filter, the overall passband characteristic will be unlike that of an individual filter and determined by the cascade of the individual characteristics. For many filter stages, the end–to–end bandwidth can be considerably less than that of an individual filter.
Different technologies define the physical properties of the OADM filter; different network applications emphasize different properties. In simple terms, OADM architectures can either be static or dynamic. The decision to deploy one architecture over another is dependent on many factors including cost, optical performance, and the degree of flexibility desired in a network.
Static OADMs typically enable adding/dropping of a specified number of channels at particular locations, providing significant cost savings in certain network scenarios. A dynamic or reconfigurable OADM architecture is more flexible than a static OADM architecture, meaning an increased ability to add/drop any and all channels at intermediate nodes. With dynamic OADM architectures, preselection of the added/dropped channels is not a requirement in the system design. In this case, switching, adding, or dropping channels simply involves reconfiguring the OADM. Dynamic OADMs are more expensive to implement and can be cost–prohibitive in many networks. A third OADM architecture has emerged to provide the flexibility diverse networks require at a price that benefits the cost/performance quotient. These flexible OADM architectures combine band and per–channel filter technologies to meet requirements of complex traffic management.
Two optical–filter technologies dominate static OADMs: dielectric thin–film interference filters and fiber Bragg gratings (FBGs). Thin–film cavity filters are structurally similar to etalons in that the mirrors surrounding the cavity have multiple layers of reflective dielectric thin–film. The layer thickness of the substrate material and the number of cavities determine the optical characteristics of the filter. Typically, the more cavities, the wider the passbands and the sharper the filter roll–offs, both very desirable filter features. The device acts as a bandpass filter, passing through a particular wavelength and reflecting all the other wavelengths. The length of the cavity determines the wavelength that passes through.
Fiber Bragg gratings operate by periodic refractive–index perturbation in the fiber. The refractive index is varied by exposing the fiber to ultraviolet light. Fiber Bragg gratings are based on the Bragg effect, in which periodic index perturbations are created to form mirrors longitudinally aligned inside the fiber core. The inscribed periodic mirrors reflect light at wavelengths that satisfy the Bragg condition, which corresponds to a wavelength twice the grating period times the effective index of refraction. Grating–based filters are characterized by low loss and high channel isolation, in addition to flat tops.
An elegant and simple OADM can be constructed using FBGs and optical circulators (see Fig. 1). In such an OADM, a reflective FBG is connected to the second port of a three–port optical circulator to create a bandpass filter from the first to the third port of the device. A second optical circulator at the other end of the Bragg grating enables optical add/drop multiplexing of one or more channels of the DWDM traffic. This type of OADM architecture has low insertion loss and high isolations of the add and drop channels.
An alternative to using optical circulators for signal separation that is particularly suitable for implementation in planar waveguides is to write a pair of identical gratings in a Mach Zehnder interferometer. The reflected signal at the Bragg wavelength is then output at the crossport. This architecture has low passband loss but poor isolation because of the variations in the 3–dB couplers from their nominal values.
Static OADM architectures are the most commonly deployed OADM because they are cost–effective and increase efficiency in terms of optical performance, especially for point–to–point traffic over longer distances.
As with static OADMs, the transmission characteristics of the tunable filter are the primary design requirement for a successful dynamic OADM. As the system becomes more flexible, optical insertion loss typically increases, or the filter's optical performance decreases compared to static filter technologies. Because of this, dynamic OADM architectures have been cost prohibitive.
From a technical standpoint, OADMs can be dynamically reconfigured by employing an array of 2 × 2 thermo–optic switches as enabling add/drop elements. This type of configuration enables any and all channels or wavelengths to be dynamically added/dropped or expressed through by simply changing the switch state. A down side to this configuration is that the number of switches increases as n2. For example, a 16 × 16 crossbar switch would need 256 2 × 2 switch elements.
Alternatively, microelectromechanical systems (MEMS) technologies can integrate arrays of optical switches in either two–dimensional (still n2) or in three–dimensional arrays (2n). These alternative switch technologies require the use of back–to–back DWDM multiplexers and demultiplexers, adding to the cost of the deployment and application. In addition, these dynamic OADM configurations generate significant optical–power loss that requires amplification (see Fig. 2).
While static OADMs do not provide the flexibility required in certain parts of the network, reconfigurable OADM architectures may prove too costly for budget–constrained service providers. For networks that demand greater variability in the volume of add/drop traffic, flexible OADM solutions are a viable alternative. Flexible OADM architectures combine band and channel filters to maximize the cost–efficiency of the former and the flexibility of the latter.
Typically, a banded filter will drop four or eight adjacent channels. This option is beneficial for point–to–point traffic and sites where a high volume of traffic terminates, such as hub sites. However, in practical networks with more than eight nodes on the network, the source and destination across the network on the day of commissioning is virtually impossible to know. In these cases, more flexibility is required. Alternatively, smaller sites may not require an entire band. In these cases, the granularity of the per–channel filter provides a cost–effective solution.
With a flexible OADM architecture, band and channel filters can be used together, cascading multiple filters between amplifiers at appropriate sites. In such applications, more than one filter will be commonly installed per location. Many locations will be populated with individual single–channel filters (not consecutive) and other larger sites may have banded filters to efficiently handle incoming traffic and allow for graceful network expansion.
Metro and long–haul/extended long–haul networks have unique requirements. In metro, the key driving factor is profitable service delivery. Service providers are interested in commissioning more services in a shorter time, and less concerned with capacity planning. Any delays can mean revenues lost to competitors.
The metro network is also extremely asymmetrical from two perspectives. First, distance between nodes on a metro ring ranges from nearly 0 to 80 km, depending on the location of the central office and point–of–presence (POP). Second, nodes on the same ring can have light traffic originating from them while others support a great deal of traffic. To support this level of diversity, metro networks generally benefit most from flexible OADM architectures.
In contrast, the key driving factor in long–haul/extended long–haul is cost efficiency. Service providers want to maximize the efficiency of capital and operating expenditures to make the most of the networking dollar. Capacity planning and management are essential, and are more easily managed given the predictable traffic volumes across the long–haul network. The long–haul/extended long–haul network is typically symmetrical compared to metro. In most cases, network elements are evenly spaced between points with equidistant node locations, and traffic flow is more predictable. For this area of the network, static OADM architectures provide an optimal and cost–effective solution.
Jim Sauer is director of product marketing at Cisco Systems, 2200 East President George Bush Turnpike, Richardson, TX 75082. He can be reached at email@example.com.