Seeing the light

Communications News, Feb, 1998 by Daniel Harman

Although you might think the greatest practical revolution in data communications recently was the development of fiber optics, the real -- if unexciting -- revolution was the almost universal adoption of Category 5 copper cabling as the basis for most data networks.

Category 5 cabling, through a simple principle of physics that reduces attenuation due to magnetism, is inexpensive, high-performing, and easy to install. But there is a serious drawback: copper is nearing the limits of its bandwidth capabilities and will require replacement in the next few years if demands for bandwidth continue to escalate.

Businesses are using faster computers, larger applications, high-quality videoconferencing, multimedia distance learning, and many forms of real-time imaging -- all band-width-eaters.

One response is to push for more client/server computing. Client/server applications can reduce network traffic by eliminating the need to move applications or great amounts of data across the wire, but they don't solve the problems of networked real-time multimedia-type applications. They also don't jive with something that we all learned about ourselves from the PC revolution: Why be efficient when we have technology that lets us be extravagant?

When fiber optics first emerged as a data communications medium a few years back as the basis for FDDI (Fiber Distributed Data Interface), it represented a quantum leap in bandwidth capacity over copper wiring (from 10-16 Mbps to 100 Mbps). ATM switching has increased the bandwidth of fiber to the 2.5-10 Gbps level. Technologies like wavelength division multiplexing (WDM) promise aggregate bandwidth up to 4 Tbps (4,000,000 Mbps).

Optical fibers combine low loss with high bandwidth, which means they not only support high-speed signals, but work over long distances as well. This translates to high information capacity. In addition, many signals can be multiplexed onto a single fiber optic cable.

One of the first applications for fiber optic cables was in networks where conventional copper links wouldn't work because of electromagnetic interference. Fiber optic cables are immune to EMI and can be run safely alongside power lines.

A bonus that comes with being immune to EMI is that a fiber optic cable is also very secure. The magnetic fields that radiate around a copper cable fluctuate when changes in the signal current take place. These fields can be tapped and used for eavesdropping without cutting the cable. If eavesdroppers wanted to tap into a fiber optic network, the would have to cut the cable, causing a large drop in transmission and making it easy for the tap to be detected. This is why many government-secure links use fiber optics.

A catch-phrase often used today when referring to fiber optic networks is "Fiber to the desktop" -- using fiber optics for workstation connections and not exclusively for backbones.

Why bring fiber to the desktop? Although some companies or organizations will install fiber because of its immunity to EMI or its security, the real answer is that most of the demand for bandwidth comes from the desktop. Videoconferencing, for example, doesn't do much good on the backbone if you can't deliver it to workstations. And unless a scientific breakthrough finds a way to send huge amounts of data through copper pairs, the next great cabling revolution will involve fiber optics.

Traditionally, bringing fiber to a workstation was an expensive proposition for a simple reason: Workstations are electronic devices, not optical devices. A converter had to be used at every workstation to convert electronic signals to optical. Fiber optical signals could not be divided easily the way that electrical signals could. Electrical signals are usually delivered as voltages that can be tapped from a wire. Optical signals are delivered as power, which limits the number of workstations that can be driven by a single optical source.

It used to require a lot of converters in conjunction with traditional networking equipment to bring fiber to the desktop -- but that is changing. Converters are still needed to change electronic signals to optical ones, but network adapter cards, hubs, and switches are available now specifically for use in pure fiber environments.

All of the pieces exist today to install a fiber network. It is not my intent to push Ethernet, FDDI, ATM, or token ring. Fiber optic cabling has similarities that exist in all of these product groups, so fiber could be retrofitted to an existing token ring network today and used for Gigabit Ethernet or ATM down the road.

DISPELLING THE MYTHS

Fiber optic technology remains burdened with a few misconceptions. One is the belief that it is too expensive. Another is that it is too difficult to install. It's important to remember that the cost of any system is affected by its usage, and indeed the cost of fiber equipment is gradually coming down as usage increases.

The true cost-effectiveness of installing fiber optics would have to be weighed against the life-cycle of the cabling plant and network reliability. Most people are hoping that they will be able to upgrade their copper cabling to meet new demands, but only fiber promises to protect their investment for many years to come. Fiber figures prominently in virtually every high-bandwidth proposal currently on the books. To make a point, you might ask yourself how many times you have upgraded your copper cabling in the past 10 years.