WDM passive optical networks enable local access evolution

WDM passive optical networks enable local access evolution

Employing a wavelength-division multiplexed, passive optical network as a future-proof fiber plant minimizes opportunity costs by permitting a broadcast network in the initial build without sacrificing future point-to-point performance

patrick p. iannone and

nicholas j. frigo

AT&T research

The most prevalent forms of telecommunications to the home--telephony and cable TV--have been relatively stable for decades, with technological improvements continually and incrementally increasing the economic efficiency of delivery. However, recent advances in

coding, the popularity and power of personal computers, the growth of the Internet and the development of user-friendly World Wide Web browsers are conspiring to blur the distinctions among data, video and telephony services. The much-heralded communications convergence is in sight.

Existing networks may prove inadequate in the future. In telephony and cable-TV networks, delivered services are almost complementary in bandwidth, mode and rate symmetry properties. That is, telephony provides a low bandwidth (64 kbits/sec), switched mode (each user gets individual information), and rate-symmetric service (the same bandwidth in both directions). In contrast, cable-TV networks deliver a high bandwidth (400 to 700 MHz), broadcast mode (each user receives the same information) and asymmetric service (the bandwidth "downstream" to the user is much greater than the bandwidth "upstream" from the user).

Furthermore, the introduction of satellite television delivery shows a potential market for even more entertainment bandwidth than coaxial cable can support. Additionally, the explosive growth in the number of Web subscribers points to the potential for ubiquitous commercial transactions, as well as the practical need for much higher line rates to reduce latencies.

The diversity and volume of broadcast service suggest a need for upgradable networks with much higher capacity. On the other hand, quality of service, privacy and security issues in increasingly mixed residential and small business neighborhoods suggest a demand for switched services.

Passive optical networks

Clearly, network providers must increasingly emphasize capacity, flexibility and upgradability in local access networks--the last link of the telecommunications networks. To that end, they are evaluating passive optical networks (PONs) that bring fiber optics all the way to subscriber residences in a fiber-to-the-home (Ftth) architecture.

Two exemplary structures have been proposed for Ftth distribution: (1) the conventional PON introduced by Payne and coworkers at British Telecom Research Laboratories as telephony over PON (Tpon), and (2) the wavelength-division multiplexed (WDM) networks proposed by S. Wagner and coworkers at Bellcore (see Fig. 1). Each architecture can be used for either a fiber-to-the-curb or a Ftth network, but for brevity, only the future-oriented Ftth implementation is discussed.

The power-splitting broadcast PON (see Fig. 1a) employs the most conventional architecture and has dominated in trials and installations to date. Central office equipment impresses data on and receives data from an optical fiber that is connected to an optical power-splitting star coupler. This star coupler is located in a remote node situated close to a group of subscribers and delivers the same broadcast signal to all subscribers over individual fibers. Each of these fibers terminates on a subscriber`s optical network unit, which decodes only the information on a prearranged time-division multiplexed (TDM) time slot. Similarly, with upstream time-division multiple access (Tdma), subscriber communications are transmitted only at prearranged times to avoid interfering with other subscribers.

PONs that use this basic optical architecture can differ markedly in schemes for handling up/down traffic segregation and for suppression of scattering impairments, such as optical and radio frequency segregation, subcarrier multiplexing, wavelength-division multiplexing, time-division directional duplexing ("ping-pong"), and space-division duplexing (two fibers). Essentially, this architecture constitutes an optical broadcast tree in the optical domain, with electronics terminals at each end to perform the switching operations.

The switched WDM passive optical network (see Fig. 1b) represents the passive photonic loop (PPL) version proposed in the late 1980s by Bellcore. The switching fabric is implemented with passive WDM devices, which optically multiplex (at the central office) and demultiplex (at the remote node) signals as a function of wavelength. In this way, only the optical signals intended for a given subscriber arrive at the prescribed destination, thereby increasing both privacy and optical power efficiency.

Several upstream strategies have been proposed for the passive splitter in the upstream direction, among them using (1) distinct wavelengths through a WDM splitter; (2) distinct wavelengths through a power splitter (in these two cases, wavelengths ln+1 through l2n+1 are unique to the optical network units); and (3) inexpensive Fabry-Perot lasers with unspecified wavelengths (l0) in a Tdma scheme similar to the upstream portion of the broadcast PON.

Delivery of broadband entertainment services has also been proposed in these networks, although the fragility of the amplitude-modulated vestigial sideband (Am-vsb) signal makes conventional television delivery difficult due to scattering impairments and power loss. In one adaptation of the PPL architecture, a broadcast video signal is sent downstream to subscribers over the upstream passive segment (see Fig. 1b) and is separated from the upstream laser light with a coarse wavelength-division multiplexer at the optical network unit.

Recently, by using the robust quadrature phase-shift keying (Qpsk) modulation format found in current satellite systems, researchers have demonstrated the delivery of broadcast digital video over a conventional PON. The advantage of the robust transmission format is balanced by the need for a set-top box to convert digital Qpsk signals to the analog form needed by today`s television sets.

Thus, in a broadcast architecture, PONs have been demonstrated to be capable of delivering both narrowband services that are electronically selected, or switched, at the optical network unit, as well as broadcast digital video. WDM passive optical networks have also been shown to be capable of delivering optically switched services and, with certain constraints, of simultaneously delivering broadcast digital video.

The advantages of optical fiber advocate convincingly for the installation of all-fiber-optic passive optical networks. These advantages include not only the medium itself--each fiber strand has a capacity thousands of times greater than the "ether" of free space used in wireless RF transmission--but also:

the potential for essentially unlimited upgrades,

the ability to remove power-consuming switching electronics from the field

the prospect for higher-quality operations, administration, maintenance and provisioning procedures.

While these advantages are well-known, issues of PON implementation remain unclear. For example, the optical technology is not universal: Broadcast PON components are just now reaching commercial production levels, with WDM components somewhat further behind. Furthermore, the match between technology and the demand for services makes the evaluation of broadcast versus switched architectures ambiguous for a potential network planner. That is, installing a relatively inexpensive broadcast PON could limit longer-term switched broadband upgrades and make future operations more costly. On the other hand, installing a switched WDM network might tie up investment at large opportunity cost while advanced services are being developed. Thus, the network choices are complicated by technical and market uncertainties.

Evolution to high performance

Fortunately, a network strategy can be developed for quick deployment of an all-fiber plant in the local loop that leverages recent technological innovations. This fiber network can evolve from a relatively low-cost Tpon-like broadcast architecture to a high-performance virtual point-to-point WDM network such as PPL--i.gif., a network in which the medium is shared without the possibility of contention. Successive upgrades may require changes to the peripheral equipment, but the outside fiber plant, including the remote node, remains unchanged. The fiber plant supports both broadcast-mode and broadband switched-mode services.

Because the primary network goal is to develop a future-proof access strategy, the initial low-cost architecture must have a passive outside plant capable of supporting future virtual point-to-point services, individualized upgrades and broadcast services (see Fig. 2). This proposed architecture is logically equivalent to a broadcast PON, yet has a wavelength-dependent routing device at the remote node that can support eventual upgrades to provide virtual point-to-point connectivity.

The resultant network--called a bidirectional optical spectral-slicing PON-- employs inexpensive light-emitting diode (LED) sources at both the central office and subscribers` optical network units. The waveguide-grating router (WGR) at the remote node is more than just a dense wavelength-division multiplexer/demultiplexer: its specially designed periodicity property ensures that when the input optical spectral width is large relative to the free spectral range of the router, equal optical powers emerge from each output port on a series of wavelengths.

Thus, the use of broadband optical sources, such as inexpensive LED transmitters, causes the WGR to operate with the functionality of a passive splitter. That is, in the downstream direction, the modulated broadband optical output of the central office LED is "spectrally sliced" through the WGR with distinct spectral components, all modulated with the same information, emerging from each output port. In the upstream direction, the broadband optical output of each subscriber`s LED is combined onto the feeder fiber via the same process, with an equal fraction of each output directed toward the central office receiver.

The bidirectional spectral-slicing PON can, therefore, take advantage of conventional TDM/tdma protocols that have been developed for commercially deployed bus architectures such as Tpon. Experimental demonstrations of this architecture indicate that aggregate baseband data rates on the order of 50 Mbits/sec are feasible. These rates are more than adequate to support near-term broadband services and telephony.

To handle video services, the spectral-slicing PON is upgraded to deliver broadcast digital television in addition to data and telephony services (see Fig. 2b). Again, an LED transmitter with a wide spectral output provides broadcast services over the wavelength-dependent infrastructure. This digital network has been demonstrated with 79 Moving Picture Experts Group (Mpeg) video channels multiplexed into 16 Qpsk subcarriers over a 500-MH¥RF band. Such a broad electrical spectrum renders the video service more susceptible to power budget and chromatic dispersion constraints than the baseband service. For these reasons, a 1.3-micron LED is used for the television transmitter.

LEDs operating in the 1.3-micron band generally have higher output powers than their 1.5-micron counterparts, and their outputs are much less affected by chromatic dispersion in standard optical fiber. The outputs from the two transmitters are combined in the central office and then routed onto the feeder fiber with a coarse wavelength-division multiplexer. Similarly, a coarse WDM is employed in each subscriber`s optical network unit to segregate the two services. This approach to providing a broadcast service has the advantage that a separate hybrid fiber/coaxial-cable network overlay is not required, because the robust digital modulation format is tolerant of relatively low signal-to-noise ratios at the receiver. In addition to a standard television set at the subscriber`s premise, a set-top box is also required to convert the Mpeg-encoded Qpsk subcarriers to Am-vsb signals.

Of course, the low-cost solutions shown in Fig. 2 have some limitations. First, employing a bus architecture for the telephony-like services sacrifices many of the privacy, security, and provisioning advantages of traditional switched-mode telephone networks. Second, optical power budget and chromatic dispersion penalties associated with both the broadcast architecture and the use of relatively low-power, broadband LED transmitters restrict the data rates.

The most compelling aspects of the spectral-slicing PON with broadcast-TV capability involve cost. The network (see Fig. 2b), which is capable of satisfying near-term demand for broadband switched-mode and broadcast-mode services, employs low-cost commercially available components at the subscribers` optical network units. When demand for bandwidth outstrips the capabilities of the network, it can be upgraded to a virtual point-to-point architecture without altering the WGR-based outside plant.

Various high-performance WDM networks, such as S. Wagner`s PPL, M. Zirngibl`s Lar-net, or AT&T Research`s Rite-net, could be brought online by changing equipment at the central office and at the optical network units.

In the Rite-net architecture (see Fig. 3) a single, multiwavelength laser transmitter located at the central office is shared by all subscribers for both downstream and upstream high-data-rate switched-mode communications. The discrete laser wavelengths are matched to those of the WGR, such that each wavelength is used to establish a secure communication channel with a distinct subscriber.

At the subscriber`s optical network unit, the downstream optical signals are divided by a passive splitter, with a portion being received as downstream data, while the remainder is overmodulated with upstream information and directed back through the remote node to the central office on a separate fiber to avoid scattering impairments.

This "loopback" concept of a per- subscriber network-provided "optical chalkboard" for upstream communications not only obviates the need for expensive wavelength-registered single-frequency lasers at the optical network unit, but also improves network security. A given subscriber is incapable of interfering with another subscriber`s communications by unintentionally transmitting at the wrong wavelength.

The possibility of explosive growth dictates that the access network be continuously upgradable. However, the uncertainty in the rate of growth argues for a more evolutionary approach. Installing an inexpensive PON with a WDM infrastructure, for which the peripheral components at the central office and at subscribers` optical network units can be replaced as needed, minimizes opportunity costs in the initial build without sacrificing future performance. The fiber plant is virtually future-proof, and the peripheral upgrades can be paid for as new services are demanded. u

Patrick P. Iannone is a member of technical staff and Nicholas J. Frigo is a distinguished member of technical staff at AT&T Research at Crawford Hill Laboratory in Holmdel, NJ.