All-optical crossconnects make it to the field
P.A. Perrier, M. Chbat, and A. Jourdan Alcatel
T. Olsen Telenor
P. M. Kjeldsen Tele Danmark
B. Landousies France Telecom
D. Vercauteren Belgacom SA
The dramatic increase in traffic demand associated with new applications like online services, mobile telephony, and future multimedia services is triggering a dramatic growth in capacity requirements for medium- and long-haul transport networks. Most network providers are turning to dense wavelength-division multiplexing (dwdm) to solve the capacity problem. dwdm offers the potential of an enormous increase in transmission throughput by capitalizing on the very large bandwidth of optical fibers (several tens of terahertz).
Yet capacity is only part of the problem for today`s network; inefficiency is the other. Routing functions, such as add/drop multiplexing and crossconnection, are still performed at relatively low speeds that would make the number, size, and therefore cost of these routing nodes quite unrealistic, if the aforementioned transmission capacities are to be deployed network-wide. To avoid an explosion in the cost of the routing function, it is essential to introduce a new all-optical layer that can handle high-bit-rate signals (typically 2.5 or 10 Gbits/sec) to provide provisioning and restoration.
To achieve a global solution, network providers must implement dwdm as a building block to this new layer. Optical techniques that leverage the wavelength dimension are necessary for implementing an optical layer independent of client electronic layers.
Building this optical layer requires a number of new network elements, analogous to their electronic counterparts and adapted to the specifics of optics. Research and development activities at Alcatel examined the potential of optical networking technologies and focused on the following areas:
Fixed and semi-flexible optical add/drop multiplexers, which provide permanent or controllable access to one or a few of the transmitted wavelengths.
Fiber switched optical crossconnects, which could be used for automatic fiber provisioning and network restoration.
Wavelength-switched optical crossconnects, which allow routing at the wavelength level, with or without wavelength translation.
These network elements rely on a number of optical technologies in areas like space switching, wavelength filtering and routing, and wavelength translation. Encouraged by significant progress at the research level, Alcatel partnered with a number of scientific and commercial companies to demonstrate the feasibility of a large-scale all-optical layer in laboratory studies and on the installed infrastructures of four major national telecommunications-network providers. The company also serves as coordinator of the European consortium known as the Optical Pan-European Network (open) project. This 3-year project, part of the European Commission-funded Advanced Communications Technologies and Services (acts) research program, has successfully demonstrated the cascade operation of four all-optical crossconnects along with transmission distances exceeding 1000 km on various types of installed fiber.
The acts open consortium consists of representatives from five European network providers: France Telecom/cnet (France), Belgacom (Belgium), Telenor (Norway), Tele Danmark (Denmark), and cselt (Italy). It also includes equipment manufacturer Alcatel and four European universities: University of Essex (UK), Technical University of Denmark (Denmark), University of Gent (Belgium), and eth-Zurich (Switzerland). They share the exciting goal of interconnecting major European cities via a mesh of high-capacity optical fiber links crossconnected at transparent photonic nodes. The project uses wdm for both transmission and routing. The minimum data rate supported in the resulting network is 2.5 Gbits/sec, and its main element is a strictly non-blocking wavelength-translating optical crossconnect (wt-oxc) prototype.
This wt-oxc, which is based on all-optical technology, is being tested and its performance assessed in two separate field trials that exploit the installed fiber plant of four of the participating network providers.
The node architecture is of the "broadcast and select" type and is strictly nonblocking (see Fig. 1). Any input wavelength on any of the input ports can be routed to any output port on any desired wavelength. This architecture has intrinsic multicast capability and does not require large space switches.
Switching is performed in three successive stages. The first is a space-switching stage, where the whole wavelength multiplex from an input fiber is selected among all incoming multiplexes according to the reconfigurable routing table of the output ports. In the second stage, wavelength selection, the desired wavelength is extracted from the selected multiplex according to the reconfigurable routing table. The third stage is wavelength translation. Here, this wavelength is converted to a fixed output wavelength.
The systematic use of wavelength converters affects the performance of the node in a number of ways. In particular, wavelength translation makes the architecture strictly nonblocking. It helps optimize resource use (about 10% to 30% gain, depending on the network connectivity, the protection scheme, and the traffic matrix) and facilitates both resource allocation and network scaleability. Also, the partial signal regeneration associated with the wavelength-conversion process improves the physical performance of the node and how much the network can grow.
The wt-oxc incorporates optical components developed by Alcatel, based on InP semiconductor optical amplifier (soa) technology. For example, the space-switching stage includes arrays of optical gates operating in a multiwavelength regime. The wavelength-translation stage encompasses all-optical wavelength converters integrating soas within an interferometric structure. The nonlinear response of these devices results in partial pulse amplitude reshaping properties, which enhance the system`s cascadeability and robustness.
Implementation and performance
A wt-oxc prototype is configured as a 4 ¥ 4 ¥ 4 optical crossconnect system. It features four input/output ports with four wavelengths per port and a data-rate of 2.5 Gbits/sec per wavelength. Shared equipment protection (1:N) is provided through a tunable backup wavelength distinct from the four wavelengths used in the working multiplex. This scheme protects against a module failure within the crossconnect, providing a spare path at a different wavelength to ensure service continuity.
All input and output ports share a common monitoring board to provide data about each channel`s wavelength, power, and optical signal-to-noise ratio. A local equipment manager (lem) is used to control and monitor the various boards. It also provides a network manager (NM) interface, which supports the configuration-management, fault-detection, and performance-monitoring functions.
The wt-oxc prototype was lab tested in various configurations using transmission distances up to 1700 km and with up to four crossconnects in cascade. The results showed a worst-case receiver sensitivity difference of about 1.5 dB at a bit-error rate of 10-10 (for long pseudo-random bit sequences), indicating that the signal quality is practically independent of the number of nodes.
It was also found that the wt-oxc is tolerant to input signal variations. For instance, it can tolerate input power variations of more than 6 dB. Furthermore, the system can accommodate poor optical signal-to-noise ratios (osnrs) without performance degradation (gain of 5 dB on the osnr at the converter`s input) even in a cascade configuration. The reshaping capability of the all-optical wavelength converter provides this benefit.
Enabling technologies
Apart from low-speed opto-mechanical space switches, few components suitable for the implementation of transparent optical routing nodes are commercially available. However, in the laboratory, advanced components such as all-optical wavelength converters, soa gates, and flat-gain optical fiber amplifiers have been demonstrated and implemented in early experiments. These components all demonstrate satisfactory performance in terms of operational speed, crosstalk, and usable optical bandwidth. They are suitable candidates for introduction in field experiments.
The research team made two main technical choices to implement the crossconnect prototype:
Monolithic interferometric wavelength converters are deployed to allocate specific wavelengths to optical data streams (see Fig. 2). The wavelength converters used within the prototype are based on the principle of a cross-phase modulation in soas in a Mach-Zehnder interferometric arrangement. The signal data pattern is transferred to the new wavelength either in phase or out of phase, depending on the operating point of the interferometer. The Mach-Zehnder structure, which enables counter-propagation of the input and output signals, supports both optical filter-free operation and the possibility of translation to the same wavelength as the input signal.
Space switches based on clamped-gain soas (cg-soas) were designed for multiwavelength operation to split and steer the optical streams from one route to another (see photo). These devices feature polarization independence, large input power range, and flat wavelength response with low gain ripple. To obtain the proper degree of compactness and integration, designers fabricated and packaged arrayed cg-soa using fiber-ribbon pigtailing technology and yag (yttrium aluminum garnet) laser-welding assembly.
Field Trials
Two field trials were used to demonstrate that the open concept of an all-optical transport network is applicable to different existing fiber infrastructures varying in age, attenuation, amplifier spacing, and fiber type (standard singlemode and dispersion shifted). The two trials also supported different traffic types (pseudo-random data streams or Synchronous Digital Hierarchy frames) and management systems.
The first trial was held between Norway and Denmark using the infrastructures provided by Telenor and Tele Danmark. The second trial bridged France and Belgium using the infrastructures of France Telecom and Belgacom.
The first field trial linked Oslo, Norway to Thisted, Denmark. It leveraged a mix of terrestrial and submarine optical cables (see Fig. 3). A wdm line terminal in Oslo generated pseudo-random binary sequence traffic at 2.5 Gbits/sec on four channels. It fed a 350-km standard singlemode fiber-transmission link to Arendal, Norway, where the wt-oxc prototype was located. Arendal was linked to the Danish coast using two alternative routes. Each route crossed the sea using a 140-km repeaterless cable consisting of dispersion-shifted fiber.
On the eastern route, an optical add/drop multiplexer, assembled with commercial technologies, was located at Hjørring, Denmark, the final location for the submarine link where monitoring was carried out. Hjørring was linked to Thisted by 180 km of standard singlemode terrestrial fiber where a receiver terminal managed full measurements on the optical signal.
On the western route, Arendal was linked to Thisted via Kristiansund, Norway. In addition, a return link to Arendal from Kristiansund was equipped to demonstrate wt-oxc cascading. The chromatic dispersion of the terrestrial fiber was partially compensated (about 40%) using dispersion-compensating fiber. Because the average amplifier spacing was rather long (83 km), large amplifier gains were required. The entire network was controlled by a centralized network-management system, which was used to establish various configurations and measure system performance.
Five major network configurations featuring transmission distances of up to 1000 km on mixed terrestrial/submarine fibers were tested. The link included cascades of up to three wt-oxcs. The transmission quality was assessed in Thisted using linear Q-factor measurements based on bit-error-rate evaluations. For these five configurations, the Q factors ranged from 9.7 to 12 (see Fig. 4), thereby displaying an excellent transmission quality. Though the transmission wavelengths were in the vicinity of the zero-dispersion wavelength of the dispersion-shifted submarine cables, the team observed no system penalty stemming from 4-wave mixing, primarily due to the relatively large channel spacing (400 GHz).
The second trial linked Paris to Brussels over the terrestrial standard singlemode cable infrastructures of France Telecom and Belgacom (see Fig. 5). The wt-oxc was positioned in Lille, France. A 4-wavelength terminal in Paris fed a 150-km transmission link to Amiens, France, where a crossbar fiber switch was located. A triangular transmission topology between Amiens, Lille, and Brussels was established to demonstrate restoration capability and loop configurations around the Lille node. The perimeter of this triangle was about 450 km.
Although the average amplifier spacing was smaller than the first trial (64 km) and the fiber attenuation lower, the distance between the nodes was significantly longer (up to 450 km) and the losses related to dispersion compensation higher. This network configuration allowed the cascade of up to four wt-oxcs over a maximum distance of about 1700 km. The transmitters were fed with stm-16 (2.5-Gbit/sec) frames and the bit-error-rate performance was evaluated using the sdh B bytes. The trial used a remote centralized network-management system for configuration and performance analysis.
Quality measurements were performed on various configurations during the installation period (see Fig. 6). As seen in this figure, even the most demanding configuration (cascade of four wt-oxcs along with 1700-km transmission) suffered a system penalty lower than 4 dB with respect to the simple configuration linking Lille and Brussels (115 km, no wt-oxcs).
Evaluation after the trials showed that under field conditions, the advanced semiconductor technologies used in the wt-oxc had not degraded over a 9-month period, thus proving their reliability and maturity.
Future direction
The OPEN field trials demonstrate the added value of the selected wt-oxc architecture in an optical network. The results confirm the maturity and long-term reliability of all-optical wavelength converters and arrayed optical gates. These achievements are especially significant because they represent the first field evaluation of a large-scale optical network incorporating all-optical wt-oxcs.
These trials show that the vision of the all-optical layer is not just a pipe dream. An all-optical crossconnect is feasible and can provide management and restoration capability beyond the electrical layer. u
Philippe A. Perrier is product line manager and Michel Chbat is product line manager for advanced technologies at Alcatel (Richardson, TX), Amaury Jourdan is deputy manager for the photonics network unit at Alcatel Alsthom Corporate Research Center (Marcoussis, France), Torodd Olsen is a research scientist at Telenor (Kjeller, Norway), Peter M. Kjeldsen is an R & D scientist at Tele Danmark (Aarhus, Denmark), Bernard Landousies is optical communications R& D engineer at France Telecom (Lannion, France), and Danny Vercauteren is cost modeling expert at Belgacom SA (Brussels).