What the future holds for DWDM - pushing the speed, capacity and distance envelope J. J. (Cobus) Nel (M.Eng (Electron), University of Pretoria) Abstract In this paper we discuss what the future holds for Dense Wavelength Division Multiplexing (DWDM) with regards to the challenges facing it: to go faster, further and to use more wavelengths. DWDM is a technique that increases the amount of information that can be carried by optical networks by transmitting multiple wavelengths on a single fibre. DWDM enables networks to accommodate consumer demand for ever-increasing bandwidth. If one considers DWDM, there are three dimensions to optical scalability, namely an increase in the speed limit (e.g. 2.5 Gb/s, 10 Gb/s, 40 Gb/s, etc.), more lanes on the superhighway (e.g. 2, 4, 16, 32, 40, 80, 160, etc. channels) and the widening of the road by increasing the number of transmission windows and therefore bandwidth of the fibre. All of these have their own unique challenges some of which are discussed in this paper. Index Terms Capacity; Distance; DWDM; Optical networks; Speed; Wavelength division multiplexing; I. INTRODUCTION remains high - a dilemma for the industry. Several carriers have invested in 3G mobile licenses and in order to recover their investment, they will need to offer new services and build new network infrastructures. These new services will be bandwidth hungry as they integrate mobility and the Internet. The access network is seen as a bottleneck in the network. However, new technologies such as Asymmetric Digital Subscriber Line (ADSL) and Gigabit Ethernet (GigE) will overcome this problem by providing customers with broadband connections over existing copper infrastructure. This places more pressure on existing transmission infrastructures and the only way to alleviate this pressure is by increasing the capacity. In addition, many service providers are facing fibre exhaust in their networks. Laying more fibre will help but will not necessarily enable the Service Providers to provide new services or allow them to unify their transmission network at the optical layer. When Internet traffic started competing with voice traffic and in some cases exceeding it, it brought new challenges for manufacturers of long-haul transmission equipment. Most networks were built by forecasting bandwidth requirements. These requirements were determined by using classical formulas such as Poisson [1] and were calculated, in the case of the United States (U.S.), based on the presumption that an individual would only use bandwidth for 6 minutes of each hour [2]. These calculations ignored the amount of traffic generated by Internet access, faxes, multiple phone line, modems, and multimedia and in fact many individuals currently use the equivalent bandwidth of 180 minutes or more of traditional voice traffic each hour [2]. Carriers have stated that the level of Internet Protocol (IP) traffic in their networks currently exceeds the level of voice traffic, and this trend is forecast to continue, supporting the statement that IP traffic will remain the principal driver of bandwidth within the communications industry. New services, such as those provided by Application Service Provides (ASPs), will increase the need for bandwidth. However, these services will only take-off if the price of bandwidth falls and the level of service reliability Manuscript received May 31, 2002. J. J. Nel is with Marconi Communications South Africa (Pty) Ltd, Midrand, South Africa (phone: +27 (0)11 256 3400; fax: +27 (0)11 318 1749; e-mail: cobus.nel@marconi.co.za). Independent data sources Merged optical traffic Figure 1:Principle of DWDM DWDM enables optical networks to maximise the use of optical fibre cable by dividing the available transmission window into multiple frequency windows. Each of these windows carries Time Division Multiplexing (TDM) based Synchronous Optical NETwork (SONET)/Synchronous Digital Hierarchy (SDH) information or multimedia services, carried as IP over Asynchronous Transfer Mode (ATM), allowing unification at the optical layer (Figure 1). This paper will concentrate on DWDM technology and the enhancements taking place in order to increase the speed, capacity, and distance envelope. II. PUSHING THE SPEED ENVELOPE The majority of transport networks are based on SONET in North America and, in the rest of the world, on the European Telecommunications Standards Institute (ETSI) based technology - SDH. As transmission capacity
increases, we have seen the movement from STM-1/OC-3 to STM-64/OC-192. Transmission technology at STM- 256/OC-768 will be commercially available in the near future. The question is how fast can we economically go? A. Electronic Bottlenecks Because electronic devices determine the speed at which information can be presented for transmission, electronic bottlenecks places a limit on the speed at which information [3] can be presented and transferred over DWDM networks. 10 Gb/s is becoming the standard line rate for backbone networks. The reason being that it has an economic advantage over 2.5Gb/s systems in that it reduces equipment cost when equated at the physical equipment level as well as the overall life-cycle. The history of telecommunications would lead us to believe that the next de-facto transmission rate will be 40 Gb/s. There has always been a degree of scepticism about the validity of moving up the transmission speed curve but today there are systems being developed which will soon be able to offer 40 Gb/s per wavelength (Marconi is one of the companies developing this technology, [4]). It has become evident that at 40 Gb/s equipment manufacturers will face a number of challenges, especially when transmitting the 40 Gb/s signal over DWDM. One of the challenges is the physical limit imposed by the network elements, which need to process/forward these high-speed channels. Although it is still possible to design and build application-specific integrated circuits (ASICs) capable of generating traffic at 40 Gb/s, serious doubts are being raised as to whether electronics will ever be able to handle STM-1024 (160 Gb/s) or whether the cost associated with the development and manufacture of these elements will make them feasible [3]. B. Chromatic Dispersion Chromatic dispersion is the phenomenon by which different spectral components of an optical pulse travel at different velocities along the fibre. This causes a temporal broadening of the transmitted pulse, which means that the energy of a narrow pulse is spread out over a larger time interval, resulting in potential interference between pulses. This differential delay of spectral components can be evaluated by means the dispersion coefficient of the fibre [ps/(nm km)] and differs for different types of fibre as is shown in Figure 2. times lower tolerance, that is, 50 ps/nm. This translates into about 160 km over G.655 fibre at 10 Gb/s but only about 2.5 km over G.652 fibre and about 10 km over G.655 fibre at 40 Gb/s. This means that when it comes to 40 Gb/s, chromatic dispersion is much more critical than at 10 Gb/s and therefore the chromatic dispersion compensation must be very accurate. The only way to do this efficiently is to use finely adjustable devices that can be tuned to the best value link by link. Dispersion compensation is often performed by using a length of optical fibre with opposite dispersion characteristics than the fibre deployed. Alternatively dispersion compensation can be achieved by using discrete components such as Bragg gratings [11]. C. Polarisation Mode Dispersion Polarisation Mode Dispersion (PMD) occurs when different planes of light inside a fibre travel at slightly different speeds due to mainly the asymmetry of the fibreoptic strand. The asymmetry may be inherent in the fibre due to the manufacturing process or may be caused as a result of mechanical stress, as shown in Figure 3. Everett and Fiorone describes PMD as being: The statistical phenomenon resulting in a Differential Group Delay (DGD) between the two Principle States of Polarisation (PSP) and in further distortion and dispersion of the signal due to the frequency dependence of both DGD and PSP [4]. The effect is graphically illustrated in Figure 4. Perfect Oval Stress Figure 3: Cross-Sections of Optical Fibres [6] Distances Distances [km] [km] 1000 1000 900 2.5 900 2.5 Gb/s Gb/s 800 10 800 10 Gb/s Gb/s 700 700 40 40 Gb/s Gb/s 600 600 500 500 400 400 300 300 200 200 100 100 0 2 4 6 8 10 10 15 15 17 17 20 20 Dispersion Dispersion coefficient coefficient [ps/(nm [ps/(nm km)] km)] G.655 G.655 G.652 G.652 Ideal situation Faster PSP Real situation x x y z z Polarised Optical Signal Propagation axis Dispersed Optical Signal Differential group delay (DGD) Figure 2: Limits of Chromatic Dispersion Given a 10 Gb/s signal s tolerance against chromatic dispersion of 800 ps/nm, a 40 Gb/s signal will show 16- y Slower PSP Figure 4: Polarisation Mode Dispersion
The quantity of bit errors encountered over a fibre segment is directly influenced by the amount of PMD in that specific fibre optic transmission section. Some general limitations of distances caused by PMD are given in the table below [6]. - Electronic devices can be used after the receiver has converted the optical signal into an electronic signal although it is difficult to correct an optical problem at the electronic level. Table 1: Distance Chart for PMD Data Speed (Gbit/s) Fibre PMD (ps per km) Fibre Fibre Deployed Deployed Yesterday Today (1.00) (0.50) Fibre Deployed Tomorrow (0.25) 10 60 km 230 km 781 km 40 4 km 14 km 49 km Demux Compensating means Compensating means Algorithm and driver PMD Detection PMD Detection At 10 Gb/s, PMD compensation may be required only in extreme cases where the combination of bad legacy fibre and very long distances occur. 40 Gb/s optical transmission will, however require PMD compensation over most of the existing cables. PMD can be classified as The Problem, because it presents the unfriendly characteristic of varying statistically over time and frequency [4]. This characteristic also makes it difficult to measure. Sources that could contribute to PMD are: - Buried cables - normally slow variations, mainly thermal effects - Aerial cables - normally fast variations, due to both mechanical and thermal significant stresses - Road and railroad environments - occasionally fast variations (vibration noise), possible thermal stresses - Nodes and shelters - they can likely be weak points, possible significant thermal effects and human interventions (very fast variations). There are a number of ways in which the effects of PMD can be countered: - Replace old fibre This is an expensive proposition, especially for incumbents. - Limit the exposure of fibre to sources of vibration This is not always easy because most long distance routes are located close to national road networks or railway lines. The influence of heavy vehicles and train traffic are therefore significant. - Limit overhead fibre Aerial cables are exposed to higher stresses (i.e. mechanical because of the wind, thermal from emphasised temperature gradient). The original idea of eliminating PMD totally is now almost completely abandoned. The focus is now on minimising the overall distortion caused by PMD. Three compensation techniques can be followed: - Mechanical devices can be used to squeeze a portion of the fibre in order to realign phases of the optical signal although mechanical devices are more prone to failure over long durations. Figure 5: PMD Compensation - There is only really one viable solution to PMD and that is to compensate for it while the transmission is in an optical state (Figure 5). This is the solution most industry players are currently investigating. III. PUSHING THE CAPACITY ENVELOPE A. Fibre Technology Since 1983 millions of kilometres of G.652 unshifted singlemode (SMF) fibre, optimised for 1310 nm, has been deployed. SMF was designed to operate at 1310 nm with zero dispersion. Attenuation [db/km] 10 1 0.1 I 850 1300 1550 II V S-Band III C-Band DWDM 1565 IV L-Band Figure 6: Transmission Windows λ [nm] The need arose to extend the distances between regenerators and the 1550 nm window started being used due to its lower attenuation when compared to the 1310 nm window. Most DWDM systems currently make use of this window, also called the C-band (1530-1565 nm). When using the C-band with SMF, dispersion starts playing a role, especially with 10 Gb/s systems, due to the fibre s non-zero dispersion characteristic. This gave rise to a number of carriers deploying G.653 dispersion-shifted fibre (DSF). DSF was designed to operate at 1550 nm with zero dispersion. Because a small amount of dispersion is required to reduce some non-linear effects [9] carriers started deploying G.655 non-zero dispersion-shifted fibre
(NZ-DSF). NZ-DSF was designed to operate across a window of wavelengths near 1550 nm with a small amount of dispersion (Figure 7). Dispersion in Picoseconds per Kilometre -20-15 -5-0 -5-10 -15 1310 SMF +NZ-DSF S-Band C-Band L-Band 1350 1400 1450 1500 1550 1600 1650 Wavelength in Nanometres DSF -NZ-DSF Figure 7: Chromatic Dispersion Curves for SMF, DSF and NZ-DSF [9] The primary reason for this movement is the lower dispersion and therefore the cost saving due to the need for less dispersion compensation. Even with an expected higher fibre cost there is still a significant cost saving compared to SMF. With the development of STM-64 (40 Gb/s) transmission systems this cost saving will be become more apparent due to more precise dispersion compensation being required. B. Opening Additional Windows (Figure 6) There are a variety of ways to increase the capacity of a fibre. As mentioned earlier one can use higher channel density, increased transmission rates or use additional transmission windows such as the Long (L, 1565-1625 nm) or Short (S, 1450-1525 nm) bands. Today networks use equipment that operates at the STM- 64 with 16 to 40 channels. These channels are spaced, 100 GHz apart, according to standard frequencies within the C- band. Using 50 GHz spacing effectively doubles the amount of channels but at the expense of requiring more costly filters and lasers. Using NZ-DSF allows the use of the L-band and also effectively doubles the amount of channels. In the manufacturing of optical fibre there is always an amount of water that is retained in the glass fibre. This gives rise to the relatively high attenuation characteristic in the S- band. Lucent [10] succeeded in virtually eliminating water molecules in the glass fibre making the S-band usable. AllWave fibre provides an additional 100 nm bandwidth for transmission capacity. IV. PUSHING THE DISTANCE ENVELOPE With conventional optical transmission systems, optical amplifiers are used to boost the optical signals. Typical distances of between 60 and 100 km can be achieved. However, every few hundred kilometres it is necessary to fully regenerate the optical signal. This regeneration process removes the effects of noise and other transmission impairments. Converting the optical wavelengths into electronic signals and back again, involves expensive equipment and significantly contributes to the cost of DWDM deployment. A number of technologies have emerged in recent years to achieve longer-range DWDM systems. These includes amplification used in conjunction with Erbium Doped Fibre Amplifiers (EDFAs) to boost overall optical amplification, Forward Error Correction (FEC) allowing for a gain in the region of 6dB by automatic detection and correction of errors and the use of Soliton technology to increase overall distance performance. In the following section we will address in more detail Raman amplification as well as the emerging Soliton technology. A. Amplification (Figure 8) There is a drive to increase the distances between amplifier sites on DWDM links. This will result in a cost reduction due to a decrease in the number of EDFAs per optical link. Stimulated Raman Scattering (SRS) occurs when a sufficiently large pump wave is co-launched at a lower wavelength than the signal to be amplified. Amplification occurs when the pump photon gives up its energy to create a new photon at the signal wavelength [8]. TX 1550 nm SIGNAL POWER AMPLIFIED SIGNAL 1450 nm POWER PUMP TRANFER Figure 8: Raman Amplification Raman gain increases almost linearly with wavelength offset and peaks at about 100nm. A backward pumping scheme is mostly used because it offers some advantages [8]. Power fluctuations, for instance, average out because each transmitted bit will see several milliseconds of the Raman wave. The use of Raman amplifiers allows for an increase in EDFA spacing resulting in a cost saving due to less EDFAs being needed on the link and a reduction in the number of sites that needs to be supported. B. Soliton Technology A number of impairments affect how far data can travel down fibre, of which the principal ones are loss, chromatic dispersion, and non-linearity. 1) Loss A fibre-optic cable gradually attenuates the light travelling down it. The signal is absorbed by and scattered by the fibre due to the low-level impurities and imperfections remaining from the fibre manufacturing process. In a long segment, optical amplifiers are required at periodic intervals to restore the signal. The optical signal-to-noise ratio (SNR) degrades gradually along the fibre length due to an increase of the amount of noise on the optical signal. Because a minimum SNR is required for error-free transmission, the total length RX
of the system is limited. 2) Dispersion Dispersion was addressed in detail in an earlier section of the paper and refers to the broadening and overlap of the data bits in a signal as it propagates along the fibre. 3) Non-linearity The intensity of light changes the refractive index of fibre causing a phase modulation of the light as it is transmitted. This leads to a change in the optical frequency along the pulse of light, which in turn leads to pulse broadening, limiting system bandwidth. Soliton Technology uses each data bit differently a different modulation format taking the form of bell-shaped pulses of light, instead of square-shaped pulses. Bearing in mind that solitons are waves, work has been carried out for some years to use them to balance the effects of dispersion and non-linearity. In effect, in this context, two wrongs make a right. Together non-linearity and dispersion maintain the integrity of the light signal. Take away non-linearity and the pulse will begin to disperse; take away dispersion and the signal will manifest non-linearity problems, such as self-phase-modulation and four-wave mixing. Soliton technology is therefore developing ways of managing solitons so that the dispersion/non-linear effects on an optical transmission system cancel one another out. Soliton based DWDM System Table 2: Capacity Increase Increase in capacity Capacity Assume a starting capacity of 40 x 2.5 Gbit/s channels. x1 100 Gb/s Replacing 2.5 Gbit/s channels with 10 Gbit/s channels x4 400 Gb/s Decrease channels spacing from 100 GHz to 50 GHz x2 800 Gb/s Include the L-Band x2 1.6 Tb/s A number of technologies are also being deployed to increase the distances achievable without regeneration. These are Raman amplification, FEC and Soliton technology. All of these will reduce cost and allow for a cost effective consolidation of different traffic types at the optical layer. ACKNOWLEDGMENT The author thanks Jim Everett (Marconi - SDH Product Strategy) and Raoul Fiorone (Marconi - Engineering expert on chromatic dispersion and PMD issues) for sharing their experiences in developing 40 Gb/s transmission systems. The author also thanks Jeff Stern (Marconi Product Strategy Director, Marconi Solstis) for providing information with regards to Dispersion-Managed Soliton Networks). 2400+km Conventional DWDM System Figure 9: Benefits of Soliton Technology Soliton Technology is continuously being refined and a number of world-record data transmission speeds have been demonstrated during laboratory trials. A standard optical fibre has carried 10 Gb/s of data across a distance of 16,000 km, and 40 Gb/s of information has been transmitted for 1,000 km [7]. V. CONCLUSION The paper attempted to show the developments under way to push both the capacity and distance envelopes. We have seen that technology is available or under development to dramatically increase the throughput of DWDM systems (Table 2). REFERENCES [1] M. Schwartz, Telecommunications Networks Protocols, Modeling and Analysis (Book style). Addison-Wesley Publishing Company, 1988, pp. 24 30. [2] The International Engineering Consortium, Dense Wavelength Division Multiplexing (DWDM). Web Forum Tutorials, http://www.iec.org. [3] M. K. Dhodhi, S. Tariq and K. A. Saleh, Bottlenecks in next generation DWDM-based optical networks, in Computer Communications, vol. 24, pp. 1726 1733, 2001. [4] J. Everett and R. Fiorone, 40 Gb/s Networks, Marconi - Product Strategy White Paper (Unpublished), April 26, 2002. [5] M. Carlson, P. Clark, S. Chabot, A. Chandonnet, N. Cyr, S. Duquet, et al, Guide to WDM Technology Testing, EXFO Electro-Optical Engineering Inc., Quebec City, Canada, 2000. [6] The International Engineering Consortium, Polarization Mode Dispersion. Web Forum Tutorials, http://www.iec.org. [7] J. Stern, Coming Soon Dispersion-Managed Soliton Networks Sending Data Further, Faster and Wider than Ever Before, Lightwave Magazine Article, January 2001. [8] The International Engineering Consortium, Raman Amplification Design in WDM Systems. Web Forum Tutorials, http://www.iec.org. [9] C. Towery & E. A. Dowdell, Advanced Optical Fibre for Long Distance Telecommunications Networks, Presented at AMTC 2000, http://www.corningfiber.com. [10] Lucent, Lucent introduces breakthrough in optical networks, http://www.lucent.com/press/0698/980604.nsa.html, June 04, 1998. [11] M. Ibsen, M. K. Durkin, K. Ennser, M. J. Cole and R. I. Laming, Long Continuously Chirped Fiber Bragg Gratings for Compensation of Linear and 3 rd Order Dispersion, Optoelectronics Research Centre, University of Southampton, Southampton, March 2000.