Grafting WDM onto existing cable systems

Fri, 01/31/1997 - 7:00pm
Venk Mutalik, Fiberoptics Engineer, Philips Broadband Networks Inc.

With the advent of 1550 nm optically amplified systems, cable companies can now provide a point-to-point link between two headends, as well as provide multi-split options at the receiving headend, thereby reaching the subscriber node without any electrical conversion. This efficient method of information transfer is transparent to system upgrades and best preserves the system CNR.

The success of a network, however, often depends on its ability to incorporate local channel program insertion in the form of analog carriers, or digital channels (for Internet access) at the receiving headend. With 1550 nm systems, as illustrated in Figure 1, no local programming content can be added at the receiving end.


Figure 2 illustrates that the standard way of providing local access today is by electrical re-conversion at the optical transition node (OTN). Here, the supertrunk channels are optically received and converted to RF. Local channels are then added to these supertrunk RF signals, which together modulate a 1310 nm optical transmitter. The final subscriber nodes (SNs) are then fed as usual. The subscriber node always receives complete channel loading which is a combination of RF1 and RF2.


In the above scheme, local channel access is always guaranteed. This process is simple, available and reliable, but is inefficient with respect to distortion performance and is dependent on the RF bandwidth requirement. This means that it is not readily upgradable.

One alternative is to optically combine local programming content with the long-haul input at the receiving headend at different wavelengths. The combined signal can then be transmitted to the subscriber node as usual as illustrated in Figure 3.


The optical local channel insertion may be carried out using either 1310 nm DFB solutions or 1550 nm solutions, depending on the specific needs of the application. This is one of the simplest and most intuitive of wavelength division multiplexing (WDM) applications. Optical components which enable this technology are now available for analog cable TV networks.

Because WDM is an optical solution, it is independent of the electronic performance of the system. This is very important in system builds where customer demands are uncertain. Thus, WDM enables cable companies to proceed with the build as they see fit and then incorporate interactive services at a later date when they become market-proven.

WDM also has other advantages: it has the potential to reduce product count, which in turn improves system reliability; and WDMs do not need any power supply and hence do not suffer from any associated heat dissipation issues. However, the economics of WDM systems are a little hard to fathom and are primarily dictated by their intended use.

The remainder of this article will examine the theory and practice of wavelength division multiplexing techniques in modern cable TV networks. Examples comparing WDM scenarios to electrical re-conversion will be presented, along with some test results. We then suggest several WDM solutions that best suit different applications. Finally, we present the enhancements in cable TV optical network architecture using WDM, with respect to cost and performance.

Theory and experiment

WDMs are being used extensively for digital telecommunications applications, where quite often, four different wavelengths in the 1550 nm wavelength region are combined. The idea is to increase the information-carrying capacity of optical systems. However, digital systems are forgiving of noise and distortions, and what is applicable for them is not readily transferable to analog communications. With this point in mind, we now examine WDM applications in an analog environment.

An optical wavelength division multiplexer is a device which can combine two (or more) different wavelengths of light. For single-receiver analog systems, the WDM typically has an insertion loss of less than 1 dB for the two wavelengths at an isolation of 25 dB between the two ports. Good WDMs also have a 60 dB optical return loss at each arm. In theory, a higher isolation can be achieved by cascading two or more WDMs on each leg to provide progressively higher isolation. This, however, results in higher insertion loss and increases cost.

When analyzing a single receiver system, as in Figure 3, it is important to consider ALL noise contributions. Noise contributors in this case are the 1550 nm system, the 1310 nm system and the receiver system.

The total noise of a WDM system is an aggregate of the Relative Intensity Noise (RIN) of 1550 nm and 1310 nm lasers, shot noise generated from the 1550 nm and 1310 nm radiation, fiber noise generated by the 1550 nm and the 1310 nm systems and thermal noise generated independently within the receiver. The CNR of each individual carrier at 1550 nm or at 1310 nm would then simply be a ratio of the carrier power of that individual carrier to the TOTAL noise in the system at the receiver, i.e.

Individual noise contributions depend on the power of that particular wavelength incident on the receiver. Carrier power is a function of the power of a particular optical wavelength and the optical modulation depth of the transmitter.

For electrical regeneration, on the other hand, all of the light from the supertrunk (1550 nm) is converted to its electrical counterpart, additional channels are added and then the whole RF combination is fed to another (1310 nm) optical transmitter for further distribution.


Figure 4 above is a chart showing calculated "equalized" CNR for an optical WDM system and an electronically regenerated system on the same scale. Equalized CNR means that the CNR at the two wavelengths is equalized by a careful adjustment of optical level and modulation depth of both the transmitters assuming equal distortions. The channel line-up at the receiver is the conventional 77 NTSC. The X-axis of this chart is the number of channels that form the supertrunk (in this case 1550 nm). A "5" on the chart, for example, refers to five channels from the supertrunk and 72 channels on the local access. The Y-axis is the "equalized" CNR.

Based on the chart, some important conclusions follow. If you have very few channels on the supertrunk section, then it is advantageous to use an electronically repeatered system. This is because the CNR is primarily limited by the local access system.

If, on the other hand, the number of channels on the supertrunk form a larger chunk of your feed, and few channels are added on the local access end, it is best to use the WDM technology. This result is true even if the number of channels being added locally are more in number but the required CNR is significantly (about 10 dB) below that which is required for the analog supertrunk channels (the chart can be modified to incorporate any CNR differential). In either case, if you had used electronic regeneration, you would have lost between 1 dB and 1.5 dB in CNR, and an identical amount in CSO and twice that amount in CTB. This cumulative effect is what is popularly referred to as "losing 3 dB with electronic regeneration."

The concave shape of the CNR curve for optical WDM shows that the noise sources at extreme ends are dominated by just one laser, as opposed to an almost equal contribution of laser noise sources when the number of channels are equal.

Several experiments were conducted to verify cable TV performance in WDM applications. In a representative test, a 1550 nm externally modulated transmitter with 77 NTSC channels (50–550 MHz) was followed by a double-pumped EDFA. This was in turn followed by 46 km of standard singlemode fiber (SMF-28). The 1550 nm output at this, the end of 46 km, was multiplexed with a standard 1310 nm transmitter which was modulated with 13 channels (between 600 MHz and 750 MHz) using a WDM. This combination was then followed by 14 km of SMF-28 to reach a standard cable TV receiver. The receiver had -0.7 dBm of 1550 nm light and -8.2 dBm of 1310 nm light. Both the transmitters were modulated with the optimum modulation depth. The final CNR for 1550 nm NTSC channels was 50 dB at CSO and CTB of -65 dBc. The 1310 channels had a CNR of 46 dB at distortions of -60 dBc. These numbers agree well with the theory presented earlier.

In summary, if digital channels are added, optical WDM is always advantageous. However, if a vast number of local analog channels (many more than the supertrunk channels) are to be added, it is best to use electronic regeneration. This limitation must always be kept in mind. From now on, we will assume that the number of channels added locally are always small, compared to the number of channels that are from the supertrunk, unless noted otherwise.


Let us now consider other architectures involving optical WDMs.

Architecture A: 1550/1310 Single WDM/single receiver.


In this case (see Figure 5), a long-haul supertrunk connects headends, and a 1310 nm transmitter is added at the transfer node to provide local access. As long as the CNR requirements are low for local access (for example, CNR needs to be about 40 dB or less for an Internet service application), this method is the best single-step upgrade, economically speaking. Even though many receivers and WDM are shown in Figure 5, one transmitter/WDM combination is needed for each receiver. So, the architecture is a single WDM/single receiver configuration.

Architecture B: 1550/1310 Single WDM/multiple receivers.


For small and well-defined systems, it is possible to cover the whole area with just one additional digital transmitter (see Figure 6). The architecture in "A" above could be modified by splitting right out of the WDM, thereby splitting its cost among the receivers. Generally speaking, the splitter losses are relatively similar for the two wavelengths, but the fiber loss is different. For optimum performance, some care must be taken in overall system design to get the proper optical mixture of the two wavelengths to the receiver.

Architecture C: 1550/1550 Single WDM/single receiver.


For systems where high CNR is required at the end of a long link (such as a long-haul headend-to-headend consolidation), split-band designs may be modified by using an optical combination (see Figure 7). Because the EDFA optical gain bandwidth is 1530–1560 nm, this scheme would allow fiber consolidation with EDFAs. Also, this architecture has the added advantage of offering dispersion compensation in the supertrunk application, if needed. For a proper implementation, however, remember that WDM devices that can isolate closely spaced wavelengths are expensive and have a higher insertion loss (1–2 dB higher than 1310/1550 WDMs). Usually, 1552 and 1557 are the preferred wavelengths for this method.

In the OTN, either Architecture A or B could be used to incorporate local channels as illustrated in Figure 8.


Architecture D: 1310/1550 dual WDM/dual receiver.


Until now, we have only considered a single receiver configuration, but a network designer can also use a dual receiver configuration. With this method, true "doubling" of fiber capacity is realized (see Figure 9). This "increase" in fiber capacity comes at a price, however, in the form of a wavelength division demultiplexer (WDD) at the receiver end. The requirements on a WDD are much more stringent than those on a WDM. A low insertion loss and an isolation of about 45–50 dB between the ports are needed.

Consider the transmitters at 1550 nm and at 1310 nm. If the two ports at the WDD are not well-separated, some of the signal from the 1550 nm transmitter "bleeds" through to the receiver tuned to the 1310 nm transmitter. Because both transmitters are now fully loaded and have similar channel line-ups but different information, the corresponding RF carriers of the two transmitters are centered at exactly the same frequencies. For example, the carriers for Channel 12 are at 205.25 MHz for both the transmitters. When a carrier of the 1310 nm transmitter is turned off for measuring CTB, the carrier of the 1550 nm transmitter would be visible as CTB. So, the measured CTB would never be better than the isolation available between the two ports.

When the isolation between the ports is greater (45–50 dB), this means that the RF carrier of the 1550 nm transmitter is now about 90 dB below the carrier of the 1310 nm transmitter and so, true CTB measurement is possible. As discussed earlier, the stringent requirement on isolation is met by a progressive cascade of single-stage WDMs; available for a higher price and insertion loss.

In addition, EDFAs for analog communications are always operated in saturation, and they are usually operated in a constant output power mode. So, when two separate wavelengths within the gain bandwidth are input to the EDFA, the output power continues to remain the same, thereby implying that the output power of each transmitter is 3 dB below the power of the EDFA. This has significant impact in system design for the rest of the network, with respect to power budgets, and must be carefully investigated before-hand.


The power and appeal of WDM is the instant "plug and play" capability described in the above section. With a proper understanding of system components and architecture, it is possible to enhance cable TV networks and take them to the next level. As interactive services grow in volume, interest in WDM is likely to increase proportionally.

Because there are a large number of nodes in the field, cable companies try to maintain them at as low a cost as is possible. Hence, an architecture that leaves the nodes unchanged is far superior to the one that requires a WDD and a second receiver. This is the main reason that single-receiver architectures have been discussed in more detail. However, in some cases (like long links and builds where fiber counts are limited) it makes perfect sense to use the WDD and the dual-receiver concept. Also, electrical re-conversion enables cable operators to insert local advertisements in nationally syndicated programs, a facility not readily available with optical WDM, and so this factor must be taken into account in determining a cost-effective solution.

As already indicated, the cost structure of WDM systems is heavily dependent on intended use. A good WDM that meets cable TV standards (about 20–25 dB isolation) typically costs about $500. If one WDM were to be used for one receiver, as in Architecture A, the effective cost of that node is now $500 more than it was earlier. This still might be cheaper than leasing a separate fiber to accommodate interactive services or local channel insertion as sometimes required by local government ordinances.

Architecture B, on the other hand, splits the cost of the WDM among several receivers. The only drawback is that the area to be served must be well-defined, and appropriate care must be taken to insure proper optical levels in the receiver. This architecture would divide the cost of the extra transmitter among several potential interactive customers, thus making good economic sense. This, in any case, is superior to electronic repeating, both in terms of distortions and CNR.

Let us now consider Architecture C. It is generally more difficult to build an analog-grade 1552/1557 nm WDM than it is to build one at 1310/1550 nm. Because of the tight wavelength specifications, several cascades of WDM are needed to provide adequate isolation. Also, because the two wavelengths are so close to each other, drift or stray interference between the two transmitters could lead to instability.

In general, however, transmitters used for Architecture C are externally modulated and by their nature provide adequate isolation, provided the selected laser wavelengths and the WDM wavelengths are in close agreement. If there is a difference in the laser wavelengths and WDM wavelength, there will be some distortion effects resulting from the finite gain-tilt of the WDM passband, as well as CNR reduction from crosstalk. Work on these WDMs is primarily driven and subsidized by telecommunications applications, and the component costs are likely to become more reasonable in the future. An identical discussion on 1310/1310 nm WDM would apply; however, access to such WDM components is more difficult because of limited interest from the telecommunications sector.

Architecture D is a dual-receiver configuration where "true increase" in fiber capacity is realized. Of the two schemes discussed earlier, the 1550/1310 nm scheme is best because components for these are now available. A good cable TV-grade WDD with at least 45 dB isolation typically costs $700. An architecture of this type also requires a second receiver. Depending on the nature of the requirements, such an investment might be worthwhile. All of the earlier discussions of cost and complexity would still apply.

In short, as components mature, the information-carrying capacity of fibers could be increased dramatically, assuming of course that such an increase is warranted. This will enable cable operators to maintain the fiber count unhindered by the finite electronic bandwidth of present-day electronic devices.

Author Information
About the author
Venkatesh G. Mutalik is a fiber optics engineer with specialization in Erbium-doped fiber amplifiers, semiconductor optical amplifiers and fiber lasers. He spent a major part of the last two years characterizing optically amplified systems for analog applications. He has also been involved in investigating optical WDM techniques and optical components for enhancing cable TV network architecture. He can be reached via e-mail at

It is a pleasure to acknowledge the help of Rick Vogel of Amphenol Corporation-Fiber Optic Products, Lisle, Ill. for help in securing WDM and WDD components for cable TV tests.



Share This Story

You may login with either your assigned username or your e-mail address.
The password field is case sensitive.