While core and metro optical networks are the linchpin of the global telecommunications infrastructure, the architecture of the optical access network is still in development. The passive optical network (PON) is a promising way to obtain low-cost optical access. Cost is the crucial criterion because the access network must be inexpensive for small-to-medium-sized businesses and residential customers. Full Service Access Networks (FSAN), a group of 20 leading telecommunications companies including both operators and equipment makers, initiated the development of PON. Through the ITU-T, FSAN recommends and specifies the architecture, equipment, and transmission requirements for PON.
The basic architecture of PON is relatively simple (see Fig. 1). An optical-line terminal (OLT) transmits a light signal over a fiber. This optical signal, distributed to as many customers as possible by passive splitters, is converted into an electronic format by optical-network terminals (ONTs). The output of these devices provides electrical signals to customer-premises equipment (CPE). Active optoelectronic equipment is located at the sending (OLT) and receiving (ONT) ends only, while an optical distribution network (the outside fiber plant and the splitters) includes only passive components—thus, the name of the network. The PON physical layout needs no maintenance in the field, so operating costs are minimal.
The PON architecture shares the network equipment among the maximum number of customers, enabling the network operator to split the cost of installation and maintenance. The PON open architecture also allows for adding new customers as needed.
The many scenarios of PON deployment differ in the length of the copper wire delivering an electrical signal to a CPE. In the fiber-to-the-home (FTTH) scenario, an ONT is installed at the customer premises so that fiber runs almost to the customer's desk. In the fiber-to-the-cabinet (FTTCab) scenario, fiber terminates at the remote terminal and, using DSL, traffic is delivered to customers over a twisted pair. Because the transmission distance over a copper wire is very short, DSL provides enough bandwidth to cover customer needs.
From a component standpoint, OLT and ONT include both a transmitter and a receiver. A typical transceiver consists of a laser diode operating at 1550 nm (downstream) or 1300 nm (upstream) and supporting electronics. A typical receiver includes a photodiode and processing electronics. Both OLT and ONT provide O/E/O conversion as well as all the "intelligent" conversion of transmission formats.
The ITU-T specifies a single-mode optical-fiber link between OLTs and ONTs. This fiber provides the minimum attenuation and practically unlimited bandwidth. Other passive devices used in the outside plant are splitters, which split optical signals in a given ratio. Splitting an optical signal 1:4, for example, implies that each outgoing signal carries a quarter of the incoming power. The distance limitation of the PON comes into play here. To keep the cost low, no amplification or regeneration is allowed. From the transmission standpoint, an OLT (which is located either in a CO, a remote terminal, or a headend) takes all incoming traffic, reformats it, and sends it to the ONTs.
Passive optical networking is developing on the assumption that placing optical fiber closer to customers will deliver enough bandwidth for their current and future access needs. If so, it seems that we don't need to consider WDM in PON networks. However, there are several reasons that justify the use of WDM in PON.
First, coarse WDM (CWDM), using 1550 nm for downstream and 1300 nm for upstream, has been embedded into PON architecture by ITU-T recommendation G.983.1 to provide better optical isolation between channels and terminal equipment. Use of CWDM also avoids expensive beam-splitting devices in favor of well-developed WDM technology.
Second, PON's physical topology is point-to-multipoint and its logical topology is a classical broadcast. The latter means that all traffic is delivered to all the ONTs and each ONT selects its own information from the entire stream. This scenario may present security and privacy problems because, potentially, any recipient can read any message. Thus, a PON operation requires encryption of transmitted traffic. The optical solution to this problem is the use of WDM. By assigning a single wavelength or a subwavelength to an individual customer, a network operator creates a virtual private network (VPN) for every customer.
The third reason to use WDM in PON is video transmission, an attractive feature that could accelerate PON deployment. Video transmission, however, puts more pressure on the PON bandwidth. Indeed, video delivery requires from 400 to 1400 Mbit/s, depending on the number of channels and transmission formats. Transmission of high-definition TV (HDTV) will set even more stringent requirements on the PON bandwidth and, by some estimates, require 18 Mbit/s per channel. Because up to 200 channels can be transmitted, an individual customer may need as much as 3.6 Gbit/s. The PON architecture would be unable to accommodate enough bandwidth for video transmission. The most effective solution assigns a specific wavelength to video, making WDM inevitable for video transmission in PON, as stated in ITU-T recommendation G.983.3.
An optical fiber can carry all wavelengths; therefore, to build a WDM PON (WPON) we need only modify OLT and ONT. An ONT requires only an additional filter. An OLT transmitter must radiate a bundle of wavelengths. Thus, an OLT becomes a rather sophisticated optoelectronics module; however, because it serves all customers, all subscribers will share any cost increase (see Fig. 2). In a plain WPON, all wavelengths are sent to all destinations and an ONT selects its own wavelength. However, a WDM demultiplexer and a splitter can route an individual wavelength to an individual ONT.
An OLT transmitter can be implemented in a traditional WDM configuration in which each DFB laser from an array radiates at an individual wavelength. Another option is to use a powerful broadband source, such as a superluminescent LED or a fiber laser. The low cost of this equipment is particularly attractive (see Fig. 3).
A single lambda can carry from 51 Mbit/s to 2.5 Gbit/s in any transmission format. For a customer with low bandwidth requirements, a fraction of a wavelength (subwavelength) may be assigned. In this case, the transmission rate can be from 1 to 100 Mbit/s. To upgrade a WPON for such a customer, a network operator need only increase the assigned wavelength or assign an additional wavelength; thus, a customer will pay as his needs grow. Different types of ONT are used in practice for these two (full and fractional lambda) configurations.
Available WPON today can support up to 16 wavelengths. The two major versions of PON are ATM PON (APON) and Ethernet PON (EPON). ATM provides better support of legacy time-division-multiplexing (TDM) traffic, such as SONET/SDH, whereas Ethernet is a data-transport technology that inherently supports IP-oriented traffic. These features determine the difference between APON and EPON. In technical terms, Ethernet PON supports no more than 15 km at a 1:16 split or 10 km at a 1:32 split at 1.25 Gbit/s data rates; APON supports a 20-km length at a 1:32 split and 622 Mbit/s.
The PON architecture enables the network operator to deliver all the different formats (ATM, Ethernet, frame relay, and T/E) and to change a transmission format without a substantial investment. Because all the intelligence of a PON network resides at the sending and receiving ends, all it takes to run a new protocol is to upgrade the OLT and ONU equipment.
The flexibility of PON architecture enables the network to transmit voice, video, and data simultaneously. Another advantage is that PON is seamlessly integrated into the existing telecommunications infrastructure. Indeed, since OLT can accept traffic in any format, additional interfaces are unnecessary.
A potential disadvantage of PON is that its bandwidth is shared among all the customers. Thus, if PON transmits at 622 Mbit/s to 32 customers as classical APON architecture does, each customer receives about 20 Mbit/s of bandwidth. However, the PON architecture allows an operator to dynamically allocate available bandwidth to a customer as needed. This dynamic bandwidth allocation helps to resolve the potential problem of temporary bandwidth shortage that may occur with fixed bandwidth distribution.
Djafar K. Mynbaev is associate professor and telecommunications coordinator at New York City College of Technology of the City University of New York, 300 Jay Street, V 733, Brooklyn, NY 11201. He can be reached at firstname.lastname@example.org.