There is no reason to believe that the installed multimode fiber base has reached its bandwidth capacity.
With the recent advancements in networking technology, such as Gigabit Ethernet and Fibre Channel, some users have questioned the long-term viability of multimode fiber (MMF) in fiber backbones. At a superficial level, it may seem valid to raise questions. After all, neither of the two protocols will operate at the full 2-km distance that many have come to expect from MMF. As network managers make their infrastructure decisions, it's important they feel confident that their installed plant will meet both their current and future needs. This is especially important today as many users begin to extend the use of MMF into the horizontal to support higher-bandwidth protocols.
A historical perspective helps to understand this issue. Multimode fiber has gained the widest acceptance in network backbones where it has offered users the opportunity to extend link distances, increase network reliability, and lower costs by centralizing electronics. The TIA/EIA-568a Commercial Building Telecommunications Cabling Standard specifies link distances of up to 2 km for MMF backbones. The 2-km link distance is also reflected in the Fiber Distributed Data Interface standard, a favorite protocol for high-speed backbones for many years, and in Ethernet standards 10Base-FL and 100Base-FX. It should be no surprise then, that there are quite a few MMF links that cover distances up to 2 km.
Multimode fiber has worked well in these networks, delivering on the promised benefits of optical fiber: higher speeds, easy upgradability, longer distances, greater reliability, noise immunity, and others. Then came the emergence of higher-speed networks. As bandwidths increased, the distance that standard MMF could deliver decreased. For example, the maximum link lengths specified for Gigabit Ethernet are 220 m for 1000Base-SX on 62.5/125 micron 160 MHz-km fiber, and 550 m for 1000Base-LX when connected to 62.5/125-micron 500 MHz-km fiber via a mode condition launch cable. Other high-speed networking technologies such as Fibre Channel and IEEE-1394b have similar distance limitations. As we look forward to 10-Gigabit Ethernet, the distance concern over MMF becomes an even bigger issue.
So is MMF running out of bandwidth? Is it time to start thinking about a different media type? The answer to both questions is that it depends on economics. Many technologies are emerging that will help users extend the life of their installed MMF plant. In addition, new multimode fibers with greater enhanced performance are also becoming available. These new MMFs provide the advantage of supporting existing applications as well as extending coverage to 10 Gbits/sec using simple electronics. The choice as to which solution is best depends on the relative costs of these technologies in the context of the customer's particular situation. Regardless of whether it's more cost-effective to install sophisticated electronics to extend the life of installed fiber or to upgrade to enhanced MMF, these technologies continue to make optical fiber a wise investment for long-term network infrastructure, both in the backbone and in the horizontal.
All multimode-fiber solutions discussed share two common traits: They all use one wavelength of light (one frequency) and on/off keying. To understand their significance, look to the telegraph. One of the very first communications devices, the telegraph, was operated by using a single tone (frequency, which is analogous to wavelength), which turned on and off by a key. This sequence of tone/on, followed by tone/off, with varying duration, enabled the data to be sent. Most fiber-optic data solutions are not much different. The core technology is the same; a carrier frequency is turned on and off at varying duration and rates to transmit information.
The development of copper-based systems made this simple method of transmitting data virtually obsolete. For example, the telephone uses varying voltages to carry voice traffic in real time. For several years, Fast Ethernet systems have used multiple voltage levels to encode data. Basically, these systems adopted more elaborate methods to provide the needed bandwidth and distance performance. Until now, multimode fiber-optic systems have not adopted these schemes because of fiber's speed advantages. But to gain both higher bandwidth and longer link lengths on current generation (installed) multimode fiber, users may need to adopt one of two strategies.
The first approach to consider is the use of multiple wavelengths. Wavelength-division multiplexing (WDM) and dense wavelength-division multiplexing (DWDM) systems increase system capacity by simultaneously carrying multiple wavelengths through the same fiber. Widely used over singlemode fiber, WDM and DWDM are used to extend the bandwidth of both previously and newly installed networks. These systems commonly deploy eight to 16 different wavelengths. The rapid acceptance of these systems in the singlemode market has quickly driven the technology to maturity and brought cost down.
Several companies are now developing versions of this technology for multimode-fiber systems. Typically, these systems use four different wavelengths, though eight or more wavelengths are possible. System designers choose the number of wavelengths to optimize the balance between distance, bit rate, and total cost. More wavelengths increase distance or bit-rate capability, but also increase cost. The performance/cost tradeoff is also influenced by the choice of wavelength operating region, that is, whether the wavelengths are in the 850-nm or 1300-nm "window." The 1300-nm window offers higher performance on the installed base of fibers but generally at a higher cost per wavelength. By using eight different wavelengths, the effective bandwidth at any one wavelength is one-eighth of the total. By reducing the bandwidth, the distance can be increased. For example, a Gigabit Ethernet port requires a data rate of 1250 Mbits/sec. If this is split across eight wavelengths, the required data rate for each wavelength is only 156.25 Mbits/sec. It is fairly easy at that data rate to obtain a distance of 2 km if operating in the 1300-nm region. If applied to Gigabit Ethernet, such an approach could provide a full 2-km reach, allowing gigabit deployment on multimode links otherwise thought to be too long.
The second technique that can be used to enhance premises multimode systems is multilevel encoding, which uses multiple amplitudes (the equivalent of voltage levels in copper systems) to encode data. In these systems, the transmitter is driven to one of several preset intensity levels. As the number of levels increases, so does the difficulty in differentiating one level from the next. A reasonable approach would use four different levels for encoding, allowing two bits to be sent in every data symbol and effectively cutting the data rate in half. At the receiver, these different intensity levels are decoded into received data symbols. As mentioned earlier, copper-based systems have used this technique for years. Without such technology, today's modem rates would be impossible. As a point of reference, Basic Rate ISDN used a four-level encoding scheme called 2B1Q, which encodes two data bits into every symbol transmitted.
What is required to make all of this a reality? First, the development of optoelectronics, ASICs, and active electronic components has to take place. From the presentations and discussions within standards bodies like IEEE and other industry forums, there are many indications that several vendors are currently working on this. Equally critical is that this solution be affordable. Like it or not, economics continue to be a dominant factor in the adoption of new technologies. Companies are operating on tighter budgets and the cost/benefit analysis of new technologies is increasingly scrutinized.
For new installations, or those where the cost of replacing installed fiber is not prohibitive, there are many developments in multimode fibers. Specifications for Next Generation Multimode Fiber (NGMMF) are currently under development by the iso/iec jtc1/sc25/wg3 and TIA TR-42 cabling-standard committees. The current draft specifications propose a new multimode fiber that will support 10-Gbit/sec transmission using a single 850-nm laser and simple two-level encoding in links up to 300 m. This solution enables the use of short-wavelength vertical-cavity surface-emitting laser technology, which should be lower in cost than singlemode, WDM, or multilevel encoding schemes.
There is no reason to believe that the installed MMF base has reached its bandwidth capacity. WDM and multilevel encoding can keep installed MMF systems running for many years, and new generations of multimode fiber will continue to offer users the most flexible and cost-effective solution throughout the entire campus network.
Steve Stange is chair of the Fiber Optics LAN section (FOLS) of the Telecommunications Industry Association (TIA) and senior product manager for Transition Networks (Minneapolis, MN). He can be reached via e-mail at email@example.com.