Stimulus and Directions of Development

  of  New Generation Optical Networks

               In the history of technical development it is difficult to find branches in which     development has occurred as rapidly as optical fiber communication during the approach and     beginning of the third millennium. The progress in electrical communication, including Time Division Multiplexing (TDM) and electrical routing, more or less followed Moor’s law. According to this  law, the engineering of computer productivity was increased twice and the price decreased twice every 18 months. The development of optical fiber technology has exceeded this rate. The speed of transfer on each spent dollar was increased twice every 9 months, and speed of transfer on each used wavelength was doubled every 12 months. According to statistics, the throughput of optical fiber networks in 2001 was doubled every 5 months. Now the transfer of 160 channels with 10 Gbit/sec each is perceived almost as usual business. Today experimental systems  are able to transfer 80 Gbit/sec on each channel. With this speed, 160 channels can transfer 12.8 Tbit per second. That considerably exceeds the bandpass of all telephone networks existing in the world.

           From the economic point of view, extraordinary progress also is observed. The cost of    transferring one bit of information falls by a factor of two every nine months. And the cost of long distance transfer of one bit today has decreased by more than a factor of four in comparison with 1996. Thus the price of a basic element of optical networks, optical fiber (in Mbit/sec), fell   approximately 60% annually.

     What is the stimulus for the further development of optical fiber networks? 

     The introduction of the Internet dramatically changed the way in which information was handled. Today the industrial world is at the beginning of an era of ubiquitous broadband systems of communication. The growth of the traffic is mainly caused by traditional needs: by transfer of music, photo images, video, and sharing of applied files and programs. The next sharp growth of traffic led us to expect distribution of new services that require rather wide bandpass, such as broadcasting, TV, transfer of images, and video in high and superhigh resolution, etc. Such services, requiring 20 to 300 Mbit/sec of channel access, will be the reason for the increasing distribution FTTH. Already, as a result of the decrease of network costs, FTTH has become an alternative to DSL and broadband services provided via cable modem. Herewith optimized for voice, data, and video, the FTTH offers speeds of data transfer, 10/100 Ethernet, scaled to Gigabit Ethernet or higher using the standard Data Over Cable Service Interface Specification (DOCSIS) to ensure safety. According to researches of Dittberner Associates, investments by the world community in FTTX during the ten year period 2004 to 2013 will increase from $3.7 billion in 2004 to $22.8 billion in 2013. This six-time increase basically is required by Asian and Pacific Ocean regions (52.8 % investments of the world market), while the advanced regions of Europe and North America will each invest about $4 billion in 2013 by FTTX  (Figure 1).

                                                                 Figure 1.

     Annual investments on average, over the period (2004 – 2013), will be: in Europe about $300 million, North America $300 million, and Asian and Pacific regions about 1.1billion. These impressive sums will occur only in the optical access networks.

     There are other reasons for the potential growth of the traffic on optical fiber networks. It is widely known what runaway growth is occurring now in the field of cellular communication. Wireless technology is a good solution that is a profitable and quickly realizable method for connecting houses and small businesses. However, wireless is limited in distance and capacity. Also there are challenges concerning reliability and data protection. The use of optical fibers on some sectors of cellular networks solves these problems. For example, connecting a base station of a cellular network using the standard WIMAX (Worldwide Interoperability for Microwave Access) with an optical regional network using fiber is a promising solution. It is the opinion of some experts that such hybrid systems seem to be promising not only technically but also in economic terms. It is difficult to predict whether this addition (optical fiber networks) to wireless access networks will be great, since they are in a stage of a competition rather than of integration. Nevertheless it is possible to expect the occurrence of one more “source” which increases the traffic on optical fiber networks.

     Thus, fast growth of the traffic will require inevitable changes in the transport infrastructure. This change should ensure substantial growth of throughput capacity of networks, and they also should have universality, scalability, and ease of integration. This wide spectrum of requirements is now reached by use of several technologies.

     Traditional providers offering TDM, ATM and IP services usually use four-level networks: IP, ATM, SONET/SDH and WDM. This solution is not optimal.  The two-level model of a network, IP over WDM (IP by means of WDM) is considered preferable. Experts and developers consider IP over WDM as a preferred technology of the next generation of optical networks for the several reasons:

     IP technology is attractive for such properties as scalability, ability to integrate other network technologies, reliability and ease of debugging. The existence of the Internet and its role in our modern community speaks well about expediency of applying IP technology. The efforts of the network technology developers have been concentrated on improving the properties of the stack of   TCP/IP protocols. The manufacturers of the equipment actively build the products to support IP. The operators of communication services make powerful investments in creating IP-trunks. The enterprises use IP for interconnecting their division networks through global networks and for providing remote access for their employees. This domination of IP  makes it obvious that in engineering practice the infrastructure of a network should be optimized for IP.  

     On the other hand, WDM is the most promising technology. It offers huge network throughput capacity, which requires continued growth of traffic. The WDM technology becomes more and more attractive with the decrease of WDM systems cost. Based on WDM, technology develops not only in backbone networks, but also in regional, urban networks and in access networks. Furthermore, WDM networks, originally used only Peer-to-Peer networks, today correspond to the high level of requirements for flexibility of a network. An ordinary system uses a separate channel of management DCN (Data Communication Network) to transfer signals of management and control of a WDM-network. This control system, according to the concept of TMN (Telecommunication Management Network), is carried out under a centralized scheme. To realize scalability, a hierarchy of management is used. Uniting IP and WDM it is possible to use the resources of a WDM network for effectively directing IP traffic and creating a unified plane of management (apparently IP-centralized) through IP and WDM-networks. As a result, the network infrastructure appears more effective from the economic point of view and thus able to transport huge volumes of heterogeneous traffic. It includes only the transport layer (photonic) and the service layer.

                                                         All-optical Routers

      For the creation and management of unified transport services, it is necessary to have a link uniting these two network levels. For that we can consider using protocol MPlS (Multiprotocol Lambda Switching) also known as GMPLS (Generalized Multiprotocol Label Switching), based on wide application of the standard MPLS (Multiprotocol Label Switching). The MPLS technology can be used with any network protocol: IP, ATM, PPP, etc. The protocol MPLS simplifies the process of routing and increases productivity. The routing MPLS is used for the formation of virtual paths, LSP (Label Switched Path), in IP-networks. An MPLS-router, also called LSR (Label Switching Router), works like this:

     A packet, that comes from the periphery of an MPLS network is accepted by the first router. On the basis of the packet’s IP-address, the router makes a decision about the packet’s direction, determines the label's meaning (according to a class of equivalence of the packet’s progress), attaches a label to the packet, and transfers it to the following router. The following LSR determines the next step, attaches a new label to the packet and sends it to the next router. For each packet, the entrance LSR writes a label separate from the heading of the third level. This label contains the address information and information about the class of service. Therefore it is not necessary to process each packet on every intermediate router. The header of the third level of a packet is read only on the entrance and the exit of the domain. Hence, routers carry out a smaller volume of work, which means there is an increase of network productivity.

     Like switching on a label in an LSR optical router, OLSR switches radiation of different wave- lengths from entrance port to exit port. They use a wavelength as a label for creating an OLSP (Optical Label Switched Path). There are two approaches for creating a label on a wavelength. In the first approach one wavelength is used on all parts of the OLSP from the entrance router to the exit router. In the second approach the wavelength is replaced on each router along the path.

     In MPLS technology the label is assigned to each packet by the entrance router. Then the label changes on every subsequent router. As in GMPLS technology, a label (a wavelength) joins a flow of bits on the entrance router and each IP packet will be accommodated in the OLSP. The main difference between MPLS and GMPLS is that MPLS can unite labels, where two or more LSPs are combined in an intermediate router in one path, when these two (or more) LSPs follow that route; GMPLS does not allow labels to be united.

      All-optical routers (or wavelength routers) are expected to become the dominant solution in  future optical networks. In its structure it is possible to allocate three basic parts:  input/output system, optical switch matrix, and network element management system.  The circuit of one variant of an all-optical router is shown in Figure 2.

                                                                       Figure 2.

     The input/output system provides the access to the optical switch matrix through standard optical interfaces. There are two types of optical interface standards: one-wave and multiwave. One-wave standards are applied in local ports that are connected to the source of the traffic and to the equipment (ATM-switch, SONET/SDH equipment, IP-routers), processing it. The multiwave interfaces are used in ports of paths ensuring pass through of WDM multiplexers and WDM demultiplexers. The input/output system is also responsible for wavelength transformation. It can  include optical amplifiers that are required in some optical routers.

     Let's consider the wavelength transformation that plays an essential role in an optical cross-switch. If the transformation of wavelengths is absent, the same wavelength is assigned to the entire OLSP, which takes place through the network. Conflicts of wavelengths on some parts of an OLSP are possible in this case. That limits the network's efficiency. Applying transformation of wavelengths on different parts of the OLSP between routers, diferent wavelengths can be assigned. This excludes conflicts, raises efficiency of using all lengths of waves and, hence, raises efficiency of the network.

     Two types of wave converters exist: optical-electrical wave converters and all-optical wave converters. The application of optical-electrical wavelength converters in WDM cross-switching, in order to prevent crosstalk between channels, requires complex configuration that raises the price of routers and makes them less attractive. Therefore, all-optical wave converters are considered more attractive.

In all-optical wave converters the transformation is carried out by using the effect of cross-modulation of wavelengths in an active medium or by using wavelength mixing in a nonlinear medium.

     The cross-modulation of waves can be observed in a semi-conductor amplifier in two modes: cross-gain modulation and cross-phase modulation. In the first mode, the modulated optical signal of a wavelength to be converted, and a continuous optical radiation (the pump radiation) wavelength  (to transform the signal) enter the input of the semi-conductor amplifier. The modulated signal modulates the gain of the semi-conductor amplifier due to the effect of saturation. The radiation of the required wavelength becomes modulated opposite in phase of the input signal. At the output of the amplifier a filter passes this radiation and blocks the radiation with the wavelength of the input signal.

     In cross-phase modulation mode, the conversion of wavelengths is based on the phenomenon that the index of refraction of an active medium depends upon the density of the radiation, passing through the medium. The change of the index of refraction results in a change of radiation phase. The converter in a cross-phase modulation mode usually operates like an interferometer such as an asymmetric Mach-Zehnder interferometer, which has two semi-conductor amplifiers in opposing legs.

     Converters that are based on the wave mixing effect, use the product of nonlinear transformation of several optical wavelengths in a nonlinear optical medium. The intensity and the length of the output waves depend on the intensity and wavelengths of the input waves (initial radiation). These converters keep the information at about the phase and the amplitude of the initial radiation and also allow simultaneous converting of some wavelengths. Four-Wave Mixing is an example of the phenomenon of wave mixing. This category of converters also uses the generation of waves, which are a consequence of nonlinear transformation of the second power in a medium with two wavelengths: the signal and the pump. This method does not add excess noise to the signal or in the transformed signal, but requires additional attention to phasing cooperating radiations and the use of fibers with very low attenuation.

      Comparing all-optical converters of wavelengths to the requirements of the present time, the most suitable are converters that apply cross-modulation. However, more promising are the converters ensuring simultaneous transformation of several wavelengths.

     Optical switches of channels (waves of various length) can be divided into two groups:

     Switches that transform the optical signal into an electrical signal, perform the switching, and then transform the signal back to optical (O-E-O);

    Switches that are known as all-optical switches in which switching is made optically.

      All-optical switches are considered more promising for application in all-optical networks, in spite of the fact that now some of their parameters are worse than the same parameters of O-E-O switches. They are based on various technologies and effects.

      All-optical switches can be switches that:

·        use optical semi-conductor amplifiers

·        are based on the application of integrated optics

·        use thermo-optical effect

·        are electro-optical

·        are electro-mechanical

      High requirements are demanded of all-optical switches. These are the basic difficulties of creating  all-optical switches:

·        the switching time

·        the characteristics of transfer (loss, cross-talk, independence of wavelength)

the temperature stability

·        the scalability

·        the compactness (inversely proportional to the number of ports).

     Switches that are based on different technologies have essentially different limiting opportunities in which to realize their characteristics. For example, thermo-optical switches have a switching time limited to a few milliseconds. Acousto-optical and electro-mechanical switches are limited to microseconds. Switches that use integrated optics operate from milliseconds to microseconds. Electro-optical switches operate from microseconds to nanoseconds (depending on the material used; for LiNbO₃, nanoseconds). Switches in optical semi-conductor amplifiers operate in nanoseconds. Except for integrated optics technology and electro-mechanical technology (MEMS, Micro-Electro-Mechanical Switches), all others have not yet achieved sufficient potential for creating a device with many switched optical paths. Now, technical realization of MEMS with many ports is more advanced than switches using integrated optics. This technology can potentially create switching matrixes with 1000 ports. MEMS-technology also promises small losses and compactness of the switching matrix. Even though this technology does not provide realization of all requirements, MEMS seem to be the preferable candidate for application in all-optical switches.

                                          The Optical Carrier Generator

     One of the major parameters determining throughput capacity of a WDM network is the number of wavelength carriers. If traditional laser diodes with stable wavelength are used as a source of optical radiation, the quantity of OLSPs on one path is limited to approximately two hundred wavelength carriers. That is because using many lasers with different wavelengths greatly raises the system's price. Simultaneously to generating many wavelength carriers, using one source is extremely useful in the technology of optical networks. Along with significant reduction of system cost, the mutual drift of the frequency of the carriers greatly decreases. Such generators are made using supercontinuum sources.

     To generate supercontinuum radiation means forming a wide and continuous spectrum when powerful light pulses pass through a nonlinear medium. The term supercontinuum is not connected to the specific phenomenon, but rather there are a lot of nonlinear effects, which in combination generate superextension of the pulse spectrum. The nonlinear effects participating in spectral extension strongly depend on the dispersion of the nonlinear medium. Reasonable chosen dispersion can essentially lower the requirements for the power of the pulse. The widest spectrum results from the propagation of an inoculating pulse with a wavelength close to the wavelength of zero-dispersion of a nonlinear medium. The supercontinuum source of radiation gives a wide spectrum with high brightness. Such a combination of features is not given by other technologies. The spectrum of radiation of a supercontinuum source is divided into a set of narrow strips forming the carriers of optical channels. To obtain equal power for each of the optical channels, the spectrum of the supercontinuum source should be smoothed. Today supercontinuum spectrums are obtained using lasers that radiate pulses by duration in femto-, pico- and even microseconds.

      Formation a spectrum of supercontinuum is more effective on fibers with strong nonlinearity, in particular, on conical fibers and photonic crystal fibers. Crystal fibers consist of a quartz core surrounded by a cladding with a set of air micro-channels going along the fiber. The structure provides, dependent on wavelength, an effective index of refraction for the cladding and allows the fiber to pass one mode in the visible and in the near infrared spectrum. Varying the arrangement and sizes of the air micro-channels, the fiber dispersion can be adjusted over a wide range, and the effective area of the propagating mode can be adapted to increase the non-linearity of the fiber. The combination of unique dispersion and increased non-linearity can be used for effective generation of supercontinuum radiation. An example layout of a multiwave generator of optical carriers that use a supercontinuum source of radiation is presented in Figure 3.

                                                                     Figure 3.

      The radiation from the laser (1), generating powerful short pulses, is entered in a fiber (2) with strong nonlinearity.  The multibandwidth optical filter (3) is applied to the fiber’s output. The results of each output of the generator’s elements are shown schematically below. Within Figure 3, (a) illustrates radiation of the laser as a sequence of short pulses of a certain wavelength and duration, following the certain frequency, for example, f₀. At (b) it is visible. Each pulse on the output of the nonlinear fiber is transformed and represents a supercontinuum spectrum. If the pass-band of the filter (3) is chosen smaller than the frequency of the repetition pulses f₀, the radiation at the filter output (c) will consist of continuous optical carriers, whose quantity depends on the filter.

     The mutual stability of these optical carriers is much greater than those using a huge quantity of laser diodes with stabilized wavelengths, since the stability of the filter is one order of magnitude higher than that of laser diodes. This reduces the probability of errors in transfer of information or (if  needed) reduces the frequency interval between the optical carriers. That enables, in particular, an increase in the quantity of carriers. Intervals between the carriers today are up to 10 GHz, which allows the optical carrier generator to provide more than 1000 channels each with a speed of transfer of 2.5 Gbit/s.