Overcoming metro access, loss, and traffic challenges
Hurdles associated with metro DWDM include cost, pass-through loss, traffic protection, and migration activity. Solutions being developed now, such as channel banding and loss reduction schemes, promise effective metro services for the future.
Metropolitan network operators struggle with severe capacity constraints as they try to keep up with demand for greater bandwidth and diversity of services. These operators need efficient, cost-effective solutions to maximize usage and revenues within existing metro fiber infrastructures, while freeing up bandwidth for future services.
Dense wavelength-division multiplexing (DWDM) is the technology of choice for long-haul high-capacity communications, yet applying DWDM to metropolitan transport introduces different requirements, all mandatory for successful results (see Table 1). Clearly, contrasts in requirements impose different design goals on a metropolitan transmission system compared to long haul.
This comparison to long-haul DWDM equipment reveals a different set of challenges—all of which must be overcome to provision metropolitan access and services in the future. Metropolitan solutions must have multiple features. They should be compact and consume minimal power. Modular parts ease network upgrade processes and accommodate more traffic and fiber topologies. Designs must achieve low pass-through losses. Channel banding can accommodate sites with high channel termination counts. Metro solutions need to support traffic protection, whether implemented in the optical layer or the client layer. Plus, avoiding the use of an optical supervisory channel, if possible, would further minimize loss.
One immediate logistical question is whether service providers can provide floor space at current Gbit/s per-square-foot densities to match the exponential rise in bandwidth demand. The nature of the new service provider business means equipment is collocated on the premises of incumbent operators; so deploying more compact equipment can reduce expenses.
If this equipment also has low power consumption it's an added bonus.
EVERY DECIBEL COUNTS
In long haul, much longer distances with associated fiber attenuation losses mandate optical amplification. In such circumstances, less emphasis needs to be placed on the inherent transmission losses of the equipment.
In a metropolitan optical network, fiber and equipment losses are much more finely balanced. Given a ring topology and a hub-and-spoke traffic pattern, a traffic channel in a metropolitan DWDM network is likely to transit many add/drop sites serving other channels before reaching its destination, thereby putting equipment-related losses at the forefront of the distance/capacity budgets achievable.
The need to minimize all optical losses in metropolitan DWDM equipment is therefore paramount, particularly if optical amplification within the ring is to be avoided. Optical amplification significantly impacts network economics, complicates network design, and can reduce channel availability.
Reducing transmission loss by the equipment falls into two categories: minimizing the loss through equipment that must be present and removing loss elements that can be safely eliminated.
MINIMIZING EQUIPMENT LOSS
As an example of removing loss elements through better equipment design, consider a traffic-gathering ring with one central site and ten traffic-gathering sites. Each of the traffic-gathering sites is connected to the central site by a pair of diversely-routed optical channels operating at 2.5 Gbit/s. Assume the traffic-gathering sites are equally spaced around the ring.
With today's optoelectronic components, it is quite possible to achieve an end-of-life optical budget between laser and receiver of 35 dB. This budget must accommodate losses associated with inserting and extracting an optical channel from the aggregate DWDM path, the signal loss of intermediate equipment, and loss within the fiber itself. The channel experiencing the most loss in the example network originates at a traffic-gathering site closest to the central site but is routed through all the remaining sites (see Fig. 1).
Using concatenated filters at the central site reduces the path loss for the worst-case channel (single-wavelength filters always have lower add/drop losses than bulk demultiplexers such as array-waveguide components). It is also clear that the components most influencing the achievable ring circumference are the single channel add/drop filters and their loss characteristics. Such components can be optimized for through-loss for passing channels or for channel isolation (reducing the impact of incoherent crosstalk). Optimizing one parameter leads to the degradation of the other, since achieving better isolation implies adding more layers to the multilayer filter in the component, and increases its loss.
While through-loss needs to be minimized, sacrificing drop-channel isolation can undo any benefit because the sensitivity of the receiver may be degraded if a high power interfering signal is present, which can easily happen in a DWDM ring. Significant degradation to the receiver penalty implies that the power budget is squeezed, which is the exact opposite of the goal of reducing pass-through losses.
To maintain a negligible crosstalk penalty to the receiver sensitivity, it is necessary to keep the interfering signal 15 dB below the received signal at the receiver. The amount of isolation required can be determined by analyzing a worst-case scenario (see Fig. 2). In this example, 43 dB of isolation is necessary if negligible receiver sensitivity degradation is to occur.
How much influence does the value of the filter pass-through loss have? If one examines the influence on ring circumference for various add/drop loss or through-loss couplets, the need to optimize through-loss is clear (see Table 2). Low loss and high isolation can be achieved if a more complex filter design is adapted. If a second filter is added in the drop path, the isolation of the individual filters is cumulative, yet the through loss of the main path remains low.
OMISSIONS CAN MINIMIZE LOSS
Turning to what should, if possible, be omitted from the equipment design, there are two leading candidates: power monitors and multiplexing components associated with the provision of an optical supervisory channel (OSC).
Power monitors split a small percentage of the incoming and outgoing DWDM multiplex and allow the nonintrusive connection of optical test equipment. Although the percentage of power extracted is usually only 10% of the total, the insertion loss of the splitter components is typically 0.2 dB. If these are furnished at the input and output ports of every item of equipment, the impact on power budget can be significant.
An OSC is used to pass network management information from equipment to equipment. It operates on its own designated wavelength, which is extracted before and inserted after each item of equipment. Even if very low through-loss components are employed, the presence of an OSC will double the number of filters at each traffic gathering site in the example network.
The impact of monitors and OSC couplers on the scale of networks is severe (see Table 3.) A better solution is to provide the monitoring function only at the central site of traffic gathering rings; at this location, all channels traversing the ring are present, and it is the most convenient location for operations personnel to perform network health checks.
There are also alternatives to an OSC for the conveyance of management information. The management traffic can ride the optical channels themselves. It can either be borne using base band modulation of the lasers (a technique that leaves the traffic itself unaffected), or it can be embedded in the traffic through the use of an "optical frame" or extensions to, say, Gigabit Ethernet protocols. While the loss of an OSC prevents the implementation of network auto discovery, the possibility for larger networks and/or the reduced need for optical amplification may outweigh the lack of this feature.
TO BAND OR NOT TO BAND
Channel banding consists of dividing the available spectrum of channels into bands and using band filters to perform coarse WDM prior to the individual channel filters (see Fig. 3.) This has two principal advantages: migration activity, such as the addition of channels, will only affect the members of a band rather than the entire channel load, and it can lead to lower nodal losses at sites where many channels terminate.
To fully capitalize on the less intrusive capacity upgrade offered by banding, it is necessary to have a clear idea of where new traffic will appear in order to guide the allocation of bands to appropriate traffic sites. However, this is often impossible to achieve with any accuracy and a misallocation of band capacity today leads to more fraught migration in the future. Band capacity becomes exhausted at sites with unexpectedly high traffic buildup, leading to greater disruption to existing traffic as the band allocation is reworked and channels are moved between bands. Furthermore, banding does not help when a new site must be inserted into a network.
In the Figure 3 network example, banding would not be effective since each traffic-gathering site has only two filters; the addition of the banding components would not improve the through loss of sites passed. Rather, it would worsen the situation since banding components have higher insertion losses than the through loss of optimized add/drop filters. Moreover, since the channel with the highest losses is extracted first at the central site, banding at this location cannot improve the channel's reach; rather, it will reduce it.
IMPROVED BANDING SCHEMES
Better use of the optical power budget through banding occurs when there are multiple sites, each terminating significant channel counts. The following analysis of loss implications explores the trade-offs between site count, channel count, and number of wavelength bands implemented.
FIGURE 4. A traffic-gathering ring in which there are N traffic-gathering sites with W wavelengths available and B wavelength bands deployed.
The network, for this example, is a traffic-gathering ring where there are N traffic-gathering sites. The DWDM system is assumed to have W waves available, and these are used to provide W/N 1+1 channel pairs between each traffic-gathering site and the central site. The numbers of bands that are deployed are assumed to be B, where B<N. As before, the channel with the highest loss will be analyzed (see Fig. 4.)
It is assumed that concatenated band splitters provide banding. That is, if there are four bands, one component divides the band into two, and a second tier of band splitters divides each half channel into two again. If there are eight bands, up to three tiers of splitters are required. The number of band splitters needed at each band division/combination point depends on which subband is being accessed. In the analysis below, the total count has been derived manually.
Individual filters will be present in the channel's path affecting other channels served at the channel's termination site and channels terminated at other sites using the same band. The number of filters serving other channels at the same collector site is the number of channels terminated at the site less one, for example, (W/N) - 1. There are (N/B) - 1 other collector sites sharing the band, each of which will contribute 2 x W/B filters. Thus, if F is the total number of filters the channel must pass through:
F = (W/N) - 1 + ((N/B) - 1) x 2 x W/B
Given N, W, and B, then F can be determined. Furthermore, it is possible to identify a criterion by which the benefit (or otherwise) of introducing further banding to the power budget can be determined. Doubling the band count for fixed N and W will change the relative number of band splitters and individual filters in the channel's path. The loss attributable to these components is:
FB x Lf + SB x Ls
where FB is the filter count for B bands, SB the splitter count for S bands, Ls is the loss of an individual band splitter, and Lf is the through loss of a single filter unit.
If the number of bands is doubled, then the component loss changes to:
If components can be obtained that exhibit a better loss ratio than this condition, there is merit in doubling the number of bands (see Table 4).
Table 4 shows the number of filters and splitters for the worst-case path in a network of various wavelength counts, traffic gathering site counts, channel counts, and band counts. It also tabulates Ls/Lf for each doubling of the band count. For example, the second row of the table indicates that moving from one to two bands makes sense if Ls < 4Lf. The final column shows ring circumferences for a typical case when Lf is 0.5 dB and Ls is 0.75 dB. A 35 dB budget is assumed to be available, fiber loss is 0.3 dB/km, and 1.5 dB is reserved for channel insertion and for extraction.
Analysis shows that where more than eight 1+1-channel pairs are terminated at a site, banding is of advantage. In the traffic-gathering ring used as an example, the number of bands times the number of sites should remain lower than the wavelength count.
Real networks will tend to reduce the advantage of wavelength banding since it will be harder to apportion channels and sites neatly to bands. Each network must be analyzed as an individual, with the assignments of sites and channels to bands performed so as to balance channel losses and to provide capacity for perceived future growth.
For more information visit: www.cisco.com/warp/public/779/servpro/solutions/optical/docs/whatiswdm.html
Rob Batchellor is a technical marketing engineer in the Optical Transport Business Unit of Cisco Systems, 170 W Tasmin Dr., San Jose, CA 95134. He can be reached via Adrienne Low at 408-527-2082 or at firstname.lastname@example.org.