Components ready for metro networks
Emerging metropolitan networks are driving the demand for a growing portfolio of optical component products.
Cindana A. Cornwell
Nortel Networks
The telecommunications industry is no longer dominated by telephony: data traffic and the Internet are now taking over. This is good news for optical network equipment vendors and optical component suppliers focused on metropolitan network evolution, because optical networking is seen as the solution to network bandwidth constraints caused by the Internet.
Despite issues facing the world economy, the optical networking industry has not suffered a downturn. More businesses are moving to E-commerce and the Internet as a way to reduce their overall operating expense.
For these reasons, carriers are studying metropolitan network deployment as the next step in optical network evolution. In 1998, a number of wavelength-division multiplexing (WDM) metro trials were held, predominantly with competitive-access service providers in North America and Europe. Now, 1999 is forecast as the year for metro dense-WDM (DWDM) deployment. Figure 1 shows the spectral window for metro component drives.
Metropolitan component portfolio
Emerging metropolitan optical networks support a mix of Internet Protocol (IP), Asynchronous Transfer Mode (ATM), and Synchronous Optical Network/Synchronous Digital Hierarchy (sonet/ sdh) traffic. They are currently dominated by WDM and DWDM 2.5-Gbit/sec-capable sources that provide 32 to 100 channels on a fiber, with a wide variety of channel spacings (100 to 400 GHz) and link lengths (20 to 200 km).
The cost of DWDM transmitters and receivers no longer dominates network equipment costs. Competition and the advances in materials science have increased the availability of proven products from multiple vendors. Now, component choices and trade-offs made by equipment vendors and suppliers drive costs. Component choices include WDM channel spacing, filtering and modulation technologies, attenuation and amplification for reach and wavelength/bandwidth upgrades, optical add/drop multiplexing, and in the near future, optical routing and optical crossconnects.
Lasers
Indium gallium arsenide phosphide (InGaAsP) and distributed-feedback (DFB) laser structures are the material of choice. The availability of high-power DFB lasers helps minimize the signal-to-noise ratio and extend reach. In addition, there are now standards for sonet/sdh pin-outs in the 14-pin butterfly packages that allow ease of implementation of second sources.
Component suppliers have seen that cost and time-to-market issues--rather than performance--drive the decision-making process with metropolitan original equipment manufacturers (OEMs). Technically, customers prefer directly modulated stabilized laser sources at 100-GHz grid spacing or loose (unstabilized) wavelength specification at 200-GHz spacing, and they are demanding these laser sources in high volume.
Wavelength-stabilization options, whether an external grating or an internal Fabry-Perot Etalon technology, increase network reliability. Nortel Networks has developed a patented Etalon device that locks wavelengths so that they do not drift out of their acceptable range over the laser`s life (see Fig. 2). Production units have proven to be environmentally stable, and they exhibit very small wavelength shifts over extended aging periods. Providing a low-cost, compact, and upgradable solution, the stabilized laser was a key to the success of the company`s OC-192 DWDM product.
Tunable lasers can be rapidly re-tuned to different wavelengths in the International Telecommunication Union (ITU) grid, unlike existing lasers, which are each dedicated to just a single wavelength. Tunable lasers provide the telecommunications industry with an important tool for the design of new network architectures. For example, the flexibility afforded by tunable lasers could support, for the first time, wavelength-routing capabilities in the optical domain--a critical element in the development of all-optical networks.
For tributary access in metro networks, short- and intermediate-reach lasers and receivers demand component speeds currently ranging from 155 Mbits/sec to 2.5 Gbits/sec, with 10 Gbits/sec visible on the horizon. Traditionally, large commercial industrial users have driven the demand for low-speed, low-cost, tributary-access components. With the growth of new services and higher-bandwidth applications--even directly into customer premises--demands have now grown for 2.5-Gbit/sec channel speeds.
Volume and standards have set the common footprint for the 8-pin MiniDIL (dual-inline laser) module. A range of components are available in the MiniDIL format, which is aimed at markets where low purchase price and cost of use are of paramount importance. The MiniDIL uses a standard package, with a standard footprint, and requires no lead bending prior to insertion in the board. The current MiniDIL range covers bit rates from 155 Mbits/sec to 2.5 Gbits/sec.
In 1998, the first surface-mount socket solution for use with the MiniDIL range of optical products was announced. This solution allows customers to plug in standard pin-out products instead of having to use hand-soldering or surface-mounting processes. Designed to be compatible with current production-line equipment, the electrical socket is soldered onto the board and the MiniDIL is simply plugged into the socket. The socket is fully compatible with the full range of MiniDIL transmitters and receivers and causes no performance degradation at bit rates of up to 622 Mbits/sec.
This approach offers customers a wealth of advantages. It alleviates possible damage to the component, increases performance, provides the same high standards of reliability over the same operating conditions as the standard MiniDIL product, and reduces the cost of manufacture and rework. Because the use of sockets is so well characterized and established within the industry, customers will also benefit from a solution that is compatible with existing equipment and further enhances ease of use.
Nortel Networks and Hitachi (Tokyo) have tested market acceptance of the product, gaining valuable customer feedback for input into the design. Customers wanted a "pick and place" product and an automated attachment process, while avoiding the problems associated with both hand-soldering and direct surface mounting of optical components.
Use of the surface-mount socket eliminates the dirt that is left from soldering. Because dirt is kept away from the optical connect, cleaning does not need to be designed into the process. In addition, the optical train is totally maintained to minimize link loss. And because there is no need for a detachable pigtail, the MiniDIL operates over the full temperature range.
Nortel Networks is working with connector manufacturer AMP Inc. (Harrisburg, PA) to deliver increasing volumes of the surface-mount socket. Hitachi and Lucent Technologies (Murray Hill, NJ), along with Nortel Networks, last year announced their commitment to manufacture the MiniDIL to a common footprint and will also be offering a surface-mount socket solution.
Receivers
The new MiniDIL receiver is a low-cost 2.5-Gbit/sec receiver suitable for Bellcore singlemode short-reach (~2 km) applications in the 1200- to 1600-nm wavelength range. Compatible with the MiniDIL transmitter, this receiver is housed in a ceramic 8-pin MiniDIL package that uses a low-noise GaAs transimpedance amplifier. The optical train consists of a singlemode fiber illuminating an InGaAs PIN. Operation from a single 5V supply is specified as standard. Typical performance gives -24-dBm sensitivity and -3-dBm overload.
Additional receiver performance can be achieved with an avalanche photodiode (APD) pre-amp module for 2.5-Gbit/sec long-distance singlemode fiber systems. It provides -34-dBm sensitivity in an 8-pin butterfly package for 1310- or 1550-nm applications.
Multiplexers
Passive multiplexing and demultiplexing functionality is required for metropolitan applications. Because of the need for protection, high-reliability optical switches will also be key elements in the deployment of systems. Since an ever-increasing number of wavelengths are being planned, a mix of filtering technologies will play a critical role in the evolution of cost-effective architectures.
The thin-film dielectric filter approach for metro optical add/drop multiplexing at a node is a cost-effective solution for low channel count. These types of devices offer low temperature dependence, low insertion loss, and a flexible upgrade path to increase the channel count at a later date by adding further multiplexing/demultiplexing building blocks. This technology also offers a practical solution for band add/drop filters (see Fig. 3).
With the need increasing to more than 16 wavelengths at metro hubs and future add/drop nodes, arrayed waveguide gratings (AWGs) will be deployed for multiplexing and demultiplexing in a cost-effective manner. Filter spacing is a major cost driver; 100-GHz-spaced filters cost more today than 200-GHz-spaced filters.
AWGs offer low channel-to-channel loss variation and are scalable to large channel counts for very dense channel spacing. They can also provide extremely efficient use of the available bandwidth. In addition, AWGs can be integrated with other components such as tap couplers and monitors. AWG designs and products have now been proven for greater than 40 channels at 100-GHz spacing.
Because passive components are essentially bit-rate independent, the same multiplexer/demultiplexer can support either 2.5- or 10 - Gbit/sec channels.
Amplifiers
For amplification, customers will require multiple node support within the metro network, and these needs exist even in short-fiber-reach applications where the loss is from nodes and not from the fiber.
Because gain and power requirements are lower in metro than in long-haul applications, lower-cost amplifiers can be deployed. But the real trade-off for metro OEMs today is lower cost versus configurable optical add/drop multiplexers (OADMs). However, the demands for both requirements are driving novel technology solutions.
A configurable mid-stage access amplifier architecture provides the most flexibility. This solution would likely cost more to implement than even today`s bidirectional long-haul WDM amplifiers. But amplifiers with a low initial cost can be implemented today with a low-gain, low-cost module.
If architectures with low-loss fixed OADMs dominate, the need for expensive mid-stage amplifiers will lessen. Ease of implementation will be key, because amplifiers will not be needed for all metro networks, as they are in today`s long-haul networks. A full portfolio of amplifiers configured for each customer is required, with gain-flattening filters and 980- and 1480-nm pump modules for single- or mid-stage access gain blocks. Examples from such a portfolio, believed to be the world`s first semi-custom family of WDM amplifiers, is shown in Figure 4.
Attenuators and optical crossconnects
Long-haul applications have already demonstrated the need for attenuation for amplifier loss padding, channel equalization, and receiver overload control. In metro networks, attenuation is becoming a requirement for performance management and optical functionality. Variable optical attenuators have arrived on the market to meet this demand.
Meanwhile, optical crossconnects and optical routing are still in the early phases of transition from technology to products. Because of the ring, mesh, and hub architectures of metro networks, highly reliable optical switches will be key elements in the deployment of systems. We see two technologies: one for protection, and one for future large-scale optical crossconnects. For example, a "small" switch--which would include 1ٴ and 2ٴ to 8ٺ--would be used in protection, OADM, wavelength management, and small optical management (small mesh restoration). The "large" switch--32䂔 to 100𨓬--would find use in tributary side management or large optical management for large-mesh restoration and future fiber management.
The metro market
New or alternative technologies carry the risks of unknown reliability and the likelihood of time-to-market delays for full-production product. But customers have considerable choice in components and component suppliers; the real issues now are in evaluating products and managing multiple suppliers.
In the emerging high-growth metropolitan network business, which is driven by time-to-market considerations, component and system vendors must work closely together to understand and develop the best price and performance options. They must then manage technology transfer from component prototypes to product to continuous cost reduction. The major issues for volume metropolitan network deployment are technology, product availability, system lead times, and the costs of network management and service provisioning. u
Cindana A. Cornwell is a senior marketing manager at Nortel Networks, Optoelectronics Div. (Paignton, Devon, UK).