1
Evolution, Challenges and Enabling Technologies for Future WDM-Based Optical Access Networks Fu-Tai An, Kyeong Soo Kim*, Yu-Li Hsueh, Matthew Rogge, Wei-Tao Shaw and Leonid Kazovsky Photonics and Networking Research Laboratory, Stanford University Packard Bldg. Rm. 058, Stanford, CA, 94305, USA (
[email protected]) the paper. Abstract—Upgrading current-generation time division multiplexing (TDM)-based optical access networks will be a challenge in the future when end-user demand outgrows current network capacity. Wavelength division multiplexing (WDM) has been considered as an ideal solution to extend the capacity of optical networks without drastically changing the fiber infrastructure. In this paper we investigate key issues and review enabling technologies for upgrading current-generation optical access networks with WDM techniques.
I. INTRODUCTION Optical access networks draw much attention from the research community for their potential to solve the bandwidth bottleneck in the last mile. For first-generation optical access networks, the major thrust in research and development has been economical deployment, so both Academia and Industry have focused on passive optical networks (PON) using a tree topology and a media access control (MAC) protocol based on time division multiple access (TDMA). Current TDM-PON standards specify the line rate up to 1Gb/s and maximum link reach of 20 km or more. These capabilities support the highspeed broadband access needs of current residential end-users [1]. As more broadband applications appear, however, demands from end-users will soon outgrow the capacity of firstgeneration access networks. Therefore, upgrading TDM-based optical access networks will be a major challenge. WDM has been considered an ideal solution to extend the capacity of optical networks without drastically changing the currently deployed fiber infrastructure. WDM can provide a virtual point-to-point link to each end-user over a PON without complicated MAC protocols, which simplifies tasks of network management, protection, and security to the level of traditional point-to-point networks. Under the assumption that PON will be a dominant architecture for optical access, this paper reviews practical design issues and current research progress in the evolution from TDM-based optical access to a future, WDM-based solution. First, in Section II, we discuss the need for WDMbased optical access networks. In Section III, we investigate several key issues in upgrading TDM-based optical access networks with WDM techniques and review some enabling technologies. Finally, Section IV concludes our discussions in *K. S. Kim is with the Advanced System Technology, STMicroelectronics.
II. WHY WDM-BASED OPTICAL ACCESS? A. Requirements for Future Optical Access As discussed in Section I, the major goal of first-generation optical access is to provide an economic solution, which enables large-scale deployment in the field: Unlike wide area networks (WAN) or metropolitan area networks (MAN), which serve a large number of end-users, typical access networks provide services to a relatively small number of endusers. Without the benefits of large-scale cost sharing, access networks must strive to minimize cost. The cost benefits of PONs verify the assumption that the PON will be a dominant architecture for first-generation optical access networks. Once the first-generation optical access solutions have been deployed in the field, design considerations other than cost will become important in the future upgrade path. The topology, for example, need not remain a strict tree. The fiber layout may be extended as pure ring or ring plus tree architectures to provide better resiliency. Best effort, datacentric services are not likely to dominate the network bandwidth; the network must be versatile enough to provide value-added services, such as video streaming, video on demand, voice over IP, and virtual private networks (VPN), and may support storage area networks (SAN). Quality of service (QoS) guarantees will become a critical issue as well, since different services have different requirements for throughput, delay, delay variation, and bit-error rate (BER). It is practically impossible to satisfy all the QoS requirements by simple overprovisioning. Smart bandwidth allocation needs to be implemented in the network layers. All these considerations make the design of future optical access networks very challenging.
B. Comparison of WDM-PON to Point-to-Point Networks and TDM-PON In point-to-point networks, since each end-user is connected to a central office (CO) through a dedicated fiber, power budget is sufficient for very long link reach and there is no need for time-domain MAC protocols, which makes network upgrade with new protocols and higher capacity straightforward. However, overall system and maintenance cost could be prohibitive considering the amount of new fibers to be installed, the high transceiver count that is two times the
2 number of end-users, and complicated cabling inside the CO. In TDM-PON, some of the fiber and a transceiver in feeder networks are shared by end-users, and only low-cost, passive power splitters are on the light paths between the CO and endusers. This enables cost sharing of a transceiver and a large portion of fiber infrastructure and reduces maintenance cost significantly. On the other hand, MAC protocols are rather complicated, so the future upgrade of TDM-PON could be a major issue. Any change in line rate and frame format for upgrade requires a change of the MAC protocol and thereby all the equipments in the network. WDM-PON, being based on the same PON architecture, shares many benefits of TDM-PON. In addition, WDM efficiently exploits the large capacity of optical fiber without much change in infrastructure and can provide a virtual pointto-point connection to each end-user, which is totally independent of line rate and frame format. So once WDM devices have been installed, future upgrade would be as easy as in point-to-point networks. Thanks to the recent progress in the design of optical components, such as athermal arrayed waveguide grating (AWG) routers and micro-electromechanical system (MEMS) switches, the signal paths can be either completely passive or at least consume power only during switching. Table 1 summarizes the comparison of point-to-point, TDM-PON, and WDM-PON solutions. Table 1. Comparison among point-to-point, TDM-PON, and WDM-PON solutions.
Point-to-Point TDM-PON WDM-PON
Capacity Best Good Better
Cost Highest Low Higher
Upgradability Easy Difficult Easy
III. DESIGN ISSUES AND ENABLING TECHNOLOGIES FOR WDM-BASED OPTICAL ACCESS A. Frame Formats and MAC Protocols The choice of frame format and related protocol depends on several factors such as interfaces with LAN/MAN and kinds of services and applications to be supported. For example, if Resilient Packet Ring (RPR) is used in a MAN and applications in access are mostly data-centric, Ethernet is a clear choice. Currently, the following three frame formats are most popular: • • •
Asynchronous Transfer Mode (ATM) Ethernet Generic framing procedure (GFP)
Note that these define not only link layer protocol, but physical layer protocol as well. ATM is well known for its strong support of QoS; hence a natural choice for access networks where multimedia, real-time
traffic as well as non-real time data traffic are served. The major drawbacks of ATM are slower data rates and large encapsulation overhead for IP packets (also known as cell tax) compared to the other two protocols. Ethernet mainly targets business users that require datacentric services. Adopting the Ethernet protocol, service providers expect to lower system cost by using existing Ethernet chipsets and maintaining a common interface to the MAN and LAN. As for QoS, Ethernet provides class of service (CoS), a best-effort, limited form of QoS. In addition, appropriately equipped switches (albeit more expensive) can support VLANs. GFP, proposed and specified by both ANSI and ITU-T, is a simple but flexible traffic adaptation mechanism specifically designed to transport either block-coded or packet-oriented data streams over a byte-synchronous communications channel. Combined with virtual concatenation (VC) and link capacity adjustment scheme (LCAS), GFP can make SONET/SDH a flexible and elastic transport infrastructure. Readers are referred to [2] and companion papers in the same issue of the magazine for more detailed discussions on GFP. WDM is usually constructed as a fixed point-to-point link to each end-user, eliminating a need for complicated MAC protocols. If fast tunable lasers or receivers are to be used to reduce cost, however, wavelength- and time-domain MAC protocols are needed as in [3]. In such a case, proprietary frame formats and protocols will likely be implemented, initially.
B. CWDM vs. DWDM Regarding the wavelength assignment for WDM access, a designer quickly encounters the decision between coarse WDM (CWDM) [4] and dense WDM (DWDM) solutions. CWDM uses 18 wavelengths from 1270 nm to 1610 nm with a channel bandwidth of 13 nm. CWDM permits cheap components, such as athermal AWG and uncooled lasers in the network due to the wide channel spacing (20 nm): There is no need to (temperature-) stabilize laser sources or optical filters. On the other hand, DWDM achieves greater spectral efficiency and with commercially available fiber-based optical amplifiers, can provide longer reach, which makes it a better upgrade option in the long-term future. CWDM has drawn a lot of attention recently because of its cost benefit as an optical access solution. However, there are a few critical issues that must be addressed: First, unless AllWave™ fiber is used, the strong absorption at the water peak severely discourages using the channels in E-band (see Fig. 1). Second, the power budget of CWDM is relatively tight. The reason includes higher fiber attenuation, potentially larger component loss, and lack of good optical amplification. It has been proposed to use solid-state optical amplifiers (SOA) or linear optical amplifiers (LOA) to amplify the CWDM links, but the SNR degradation is much worse than that of fiberbased optical amplifiers. One of the possible network upgrade path is to use CWDM first for the cost reason. If more wavelengths are required, in-
3 service upgrade is possible by inserting DWDM channels in Cband without disturbing the existing transmission [5].
Fig. 1. CWDM specification
Spectral slicing techniques are another approach to permit lower system cost in WDM access networks. The transmitter at the ONU side uses a broadband LED instead of a laser. An AWG will slice the broadband signal and allow all data from different ONUs to be delivered to the OLT. In one such scenario, upstream traffic is multiplexed in the time domain, while downstream traffic uses WDM [9]. Another possible solution is for each ONU to have only a modulator. The optical carrier is generated at the OLT and delivered to ONUs, where it is modulated and sent back to the OLT [10]. Wavelength stability and reconfigurability are easily achieved by configuring laser source(s) at the OLT. A potential issue for this configuration is that the wave band occupied by optical carriers sent from OLT to ONUs may not be modulated with downstream data.
D. Video Broadcasting C. Light Sources for WDM-PON One key component for the successful deployment of WDM access networks is the laser diode. A laser diode can be expensive, especially if wavelength stabilization is required, as in DWDM networks. It is also very desirable to have tunable laser diodes. The tunable laser sources not only support network provisioning and reconfigurability, but also minimize the production cost and numbers in stocks for back up. Several approaches exist to achieve wavelength reconfigurability: • Tunable laser diodes, ideally tunable Vertical Cavity Surface Emitting Diodes (VCSEL). • Injection-locked Fabry-Perot lasers at ONUs. • Broadband light source, such as LED, with spectral slicing. • ONU contains only a modulator. VCSEL have the potential for low-cost mass production, since quality testing for VCSEL is much easier than DFB lasers [6]. 850nm and 1310nm VCSEL are commercially available. However, 1550nm VCSEL are not yet mature. When tunable long wavelength VCSEL technology is mature, it will be an ideal candidate for optical access networks since it has the potential for high-level system integration with backend electronics. Fabry-Perot (FP) laser sources are currently the cheap and ideal solution for PONs. However, there are two reasons for FP lasers to be less attractive in WDM networks. First, multiple longitude modes are excited in a FP laser. Intertwined with group velocity dispersion on fiber links, link reach can be limited by the mode partition noise (MPN). Second, since there is no wavelength/temperature-stabilizing unit, FP lasers cannot emit at precise wavelengths. Researches have proposed using an injection-locking mechanism to excite only one designated mode (wavelength). The FP laser in this case functions as an oscillator that synchronizes with external excitation [7,8]. Modulation index, laser bias current, and the power of external optical excitation must be carefully chosen to maximize the efficiency of injection-locked FP lasers.
Video is one important service that access networks must deliver. According to new digital TV and HDTV standards, it is possible to multiplex the digital video stream in baseband with other digital data. Considering the large amount of bandwidth needed for video data (HDTV requires more than 20 Mb/s bandwidth), however, multiplexing video signals in baseband may not be a practical option, especially for broadcast video. One alternative is using sub-carrier multiplexing (SCM) technique, which puts video channels on RF bands and can deliver more than 1 Gb/s of video signals or in other words more than hundreds of analog TV channels. Using SCM, we can overlay broadcast TV signals on the existing baseband digital data over a PON. Signals being modulated onto RF band could be 2-dimensional digital signals (e.g., QPSK- or QAM-modulated) or pure analog signals that could provide backward compatibility for existing CA-TV headends. The downstream broadcasting nature of PON makes the video distribution easy. But, there is a downside, too: The link quality to support SCM transmission on optical fiber is usually much higher than to transmit digital signals such as on-off keying (OOK). Also, due to nonlinearity of laser source and optical fiber, delivering analog video is a real challenge. As for the implementation, both frequency division multiplexing (FDM)- and WDM-based approaches are possible. In FDM-based approach, both digital baseband and RF video signals are multiplexed in electrical domain, and then modulated onto one wavelength. This approach, compared to WDM-based one, is better in terms of cost because we don’t need additional laser and photodiodes at OLT and ONUs, and electric filters are cheaper than optical counterparts. This can be used as a means to upgrade the network capacity for data transmission as well [11]. In WDM-based approach, SCM video stream is delivered on a separate wavelength at 1550 nm. In this case, the downstream digital data is at 1490 nm, while upstream digital data is at 1310 nm. Due to the wide separation between wavelengths for digital data and SCM video, the main nonlinearity that degrades the transmission quality is Raman
4 Crosstalk. The link length and signal power has to be carefully optimized to ensure good carrier-to-noise ratio (CNR) equal to or greater than 43 dB [12]. In terms of system design, this approach requires WDM filters and additional laser and photodiodes. It may not be cost-effective initially, but since video is likely to be provided by different service providers, it may be convenient to assign a separate wavelength for video service in WDM networks from the management aspect. When AWG is used in WDM-PON, we cannot exploit the broadcasting nature of PON in downstream because the output ports of AWG can pass a specific wavelength only. In [13], to broadcast video signal over WDM-PON, the authors proposed splitting the optical power of video signal into the N-1 ports of N×N AWG, while the remaining port is dedicated to digital baseband signal. With this configuration, each ONU can receive its own dedicated baseband digital channel with broadcasted video signal.
E. Protection and Restoration Protection and restoration for optical access networks is still a fairly un-developed field, since the number of users served by the network is relatively limited compared to WAN or MAN. The specifications for protection and restoration are not clearly defined yet. However, if the scale of access networks continues to grow and business users demand such qualities, protection and restoration will be indispensable. One way to protect a tree-topology PON is to connect ONUs with fiber links. The idea is to recover from a fiber cut between a remote node (RN) and an ONU. In [14], the authors have proposed grouping two ONUs for a 1:1 protection. Each ONU has its own unique wavelengths for upstream and downstream. When a fiber is cut, the optical switches inside an adjacent ONU are reconfigured such that the light path can propagate through the supporting ONU. Other research proposed a duplicated structure to protect the feeder section, which is from OLT to RN, and amplifiers on the links, for SuperPON [15]. These protection schemes naturally support existing PON architectures. In a ring topology, the access network likely encompasses more end users and may be categorized as a metro/access network. In [16], the authors proposed a simple and elegant way for restoration in a single fiber bi-directional access ring. During a fiber cut, a 1x2 optical switch is flipped to select the data from the other direction. In [17], the authors proposed network protection with protection switches in the ONRAMP project.
IV. CONCLUDING REMARKS In this paper we have discussed issues including the choice of frame format and specific WDM technology, light sources for WDM-PON, video overlay and protection & restoration, which are critical in upgrading from TDM to WDM-based optical access. Still, cost is the main obstacle in the evolution toward WDM-based optical access. Given the current fast pace of
development in component technologies and the deployment of PON, however, WDM-based optical access can be deployed sooner than expected to meet the ever-growing demands for bandwidth from end-users. Although our discussions in this paper focus on optical access solutions, the real-world implementation of access networks will be a hybrid of optical fiber links, existing twisted-pair copper lines, wireless LAN, and even free-space optics. The interfaces between the different technologies and media in hybrid access create interesting research areas with challenges in size and power minimization. In particular, the units which must eventually merge this traffic onto the network must be flexible, small, cost-effective and field-deployable.
REFERENCES [1] [2]
[3]
[4] [5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
K. S. Kim, “On The Evolution of PON-Based FTTH Solutions,” (Invited Paper) Information Sciences, vol. 149/1-2, pp. 21-30, Jan. 2003. E. Hernandez-Valencia, M. Scholten, and Z. Zhu, “The Generic Framing Procedure (GFP): An Overview,” IEEE Com. Magazine, vol. 40, no. 5, pp. 63-71, MAY 2002. M. Spencer, and M. Summerfield, “WRAP: A Medium Access Control Protocol for Wavelength-Routed Passive Optical Networks,” IEEE JLT, vol. 18, no. 12, DEC 2000. ITU-T Recommendation G.694.2 (2002), Spectral grids for WDM applications: CWDM wavelength grid. R. Iannone, K. Reichmann, and L. Spiekman, “In-Service Upgrade of an Amplified 130-km Metro CWDM Transmission System Using a Single LOA with 140-nm Bandwidth,” in Proc. OFC 2003, paper ThQ3, 2003. C. Chang-Hasnain, “Progress and Prospects of Long-Wavelength VCSELs,” IEEE Opt. Com. Magazine, vol. 1, no. 1, pp. s30-s34, Apr. 2003. H. Kim, S. Kang, and C. Lee, “A Low-Cost WDM Source with an ASE Injected Fabry-Perot Semiconductor Laser,” IEEE PTL, vol. 12, no. 8, pp. 1067-1069, AUG 2000. W. Huang, C. Chan, L. Chen, et al., “An Optical Network Unit for WDM Access Networks with Downstream DPSK and Upstream Remodulated OOK Data Using Injection-Locked FP Laser,” in Proc. OFC 2003, paper TuR2, 2003. D. Jung, S. Sin, C. Lee, et al., “Wavelength-Division-Multiplexed Passive Optical Network Based on Spectrum-Slicing Techniques,” IEEE PTL, vol. 10, no. 9, pp. 1334-1336, SEP 1998. J. Kani, M. Teshima, K. Akimoto, et al., “A WDM-Based Optical Access Network for Wide-Area Gigabit Access Services,” IEEE Opt. Com. Magazine, vol. 1, no. 1, pp. s43-s48, APR 2003. N. Chand, P. Magill, S. Swaminathan, et al., “Delivery of Digital Video and Other Multimedia Services (>1 Gb/s Bandwidth) in Passband above the 155 Mb/s Baseband Services on a FTTx Full Service Access Networks,” IEEE JLT, vol. 17, no. 12, pp. 2449-2460, DEC 1999. F. Coppinger, L. Chen, and D. Piehler, “Nonlinear Raman Cross-Talk in a Video Overlay Passive Optical Networks,” in Proc. OFC 2003, paper TuR5, 2003. E. Son, K. Han, J. Kim, et al., “Bidirectional Passive Optical Network for the Transmission of WDM Channels with Digital Broadcast Video Signals,” in Proc. OFC 2002, paper ThGG109, 2002. T. Chan, C. Chan, K. Chan, et al., “A Novel WDM Passive Optical Network with Bi-directional Protection,” in Proc. SPIE vol. 4909, pp. 167-173, 2002. A. Phillips, J. Senior, R. Mercinelli, et al., “Redundancy Strategies for a High Splitting Optically Amplified Passive Optical Network,” IEEE JLT, vol. 19, no. 2, pp. 137-149, FEB 2001. F. Dorgeuille, L. Noirie, and A. Bisson, “40km Passive Optical MetroAccess Ring (POMAR) Including a Protection Scheme Based on BiDirectional Fibers,” in Proc. OFC 2003, paper ThQ4, 2003. N. Froberg, S. Henion, H. Rao, et al., “The NGI ONRAMP Test Bed: Reconfigurable WDM Technology for Next Generation Regional Access Networks,” IEEE JLT, vol. 18, no. 12, pp. 1697-1708, DEC 2000.
