Transporting Heavy Data Is Light Work

Industry pundits may differ on their predicted time scales, but they all agree that  the volume of data traffic will far surpass the present network capacity for voice-based conversations. You only have to look at the explosive growth rate of Internet usage to understand the level of demand for high-speed data communications. The most important metric related to greatly increased data traffic is network capacity or bandwidth.

Deceptively simple in appearance, hair-thin optical fiber is the high-tech key to modern wide-bandwidth data communications. The diameter of the central core in a single fiber usually is 62.5 mm (0.0025″) for multimode (MM) fiber or about 9 mm (0.00036″) for single-mode (SM) fiber. Cladding of 125 mm (0.005″) OD surrounds the core and serves to confine light either by reflecting it in stepped index fibers or refracting it in graded index fibers. An outer, buffering layer is added to protect the fiber from moisture and physical damage.

Both types of fiber operate in the infrared area of the spectrum and exhibit very low losses compared to coaxial or twisted-pair electrical wire. MM systems are used at 850 nm and 1,300 nm, and SM fibers run at 1,300 or 1,550 nm. These wavelengths represent the location of narrow bands of frequencies used to multiplex multiple data channels onto a single fiber.

The fiber bandwidth is divided this way into three regions to avoid operation near strong light-absorbing areas at 1,000 nm, 1,400 nm, and above 1,600 nm. Absorption at these wavelengths is caused by OH+ hydroxyl radicals that occur naturally as impurities.

For MM fiber, losses are about 3 dB/km at 850 nm and 1 dB/km at 1,300 nm. SM fiber has even lower losses: 0.4 dB/km at 1,300 nm and 0.3 dB at 1,550 km. In contrast, coaxial cable losses are so large that they are specified as a number of decibels/100 ft or dB/m.

Even very large diameter coaxial cable has losses far greater than those of optical fibers, and the coaxial losses are present at much lower frequencies. Twisted-pair characteristics are even worse than those of coaxial cables. There is no question that optical fibers are technically superior, and installed cost often can be cheaper than for copper cables.

Although only 0.0025″ in diameter, MM fibers are large enough that light can travel through them on many distinct paths. Because some of the paths involve multiple reflections from the cladding, at the end of a long fiber portions of the light will have traveled different distances. This results in dispersion: a pulse of light will not be as distinct when it leaves the fiber as when it entered. Uncertainty of this kind is equivalent to jitter and reduces usable bandwidth.

In contrast, the 0.00036″ dia of an SM fiber is only a few times larger than the wavelength of the light being carried, and all the light must traverse the same path. Because only one mode of propagation is allowed, SM fibers simply don’t produce the kind of dispersion seen with MM fibers.

In graded index MM fibers, the index of refraction of the core gradually decreases with radial distance from the center. Because light travels faster in glass with a low index of refraction, the portion of the light that travels farther also goes faster. Shaping the index of refraction profile in this way eliminates most of the mode-dependent dispersion effects and yields a 100x improvement in usable bandwidth over a simpler stepped-index MM fiber (Figure 1).

The numerical aperture (NA) of a fiber is defined by the highest angle at which light can enter the fiber. Because a MM fiber has a relatively large diameter, and because light entering at a high angle will be reflected or refracted back toward the center, a MM fiber can have a large NA value. It is easier to couple light into a fiber with a high NA, and this is one reason that MM cables are driven by LEDs.1 Other factors supporting the use of LEDs are cost and an LED’s lower power that corresponds to the shorter distances for which MM fibers are suited.

SM fibers, being much smaller in diameter and with a different cladding design, have a lower NA figure. They also can provide long-range transmission. For these reasons, a higher-power laser is used to drive SM fibers. The increased cost of the SM fiber and laser driver is easily offset by the reduced number of repeaters required in very long haul networks.

One SM fiber may carry a few discrete wavelengths of light, typically 1,310 nm and 1,550 nm, in wavelength division multiplexing (WDM), or many frequencies—a product with 80 channels recently was announced—in dense WDM (DWDM) applications. In an article on fiber-optic networks, Dr. Andre Girard discussed the effect of new, all-optical components on network capacity. He noted that cables will become denser with more than 800 fibers per cable to support a predicted doubling of network bandwidth requirements every six months.

He also considered the practical bandwidth that can be provided using a single fiber, concluding that 1 Tb/s should soon be practical. One way to provide 1 Tb/s is to combine 400 wavelengths each modulated at a 2.5 Gb/s data rate.2

One problem standing in the way of very high data rates is the imperfections of the optical fibers themselves. At the high drive powers required to achieve long transmission distances, nonlinearities can limit the extent to which DWDM can be used. An example of an improved-performance fiber is the large effective area fiber developed by Corning LEAF® optical fiber.

By designing the fiber to have two separated, concentric areas of high-refractive index, higher power can be accommodated at lower light intensity. This leads to lower distortion, especially the so-called four-wave mixing (FWM) effect in DWDM applications.

In addition to the center of the core, the outer ring of high-index material also carries significant power. Distributing the light within the core in this manner results in an effective fiber area typically 32% larger than conventional fibers (Figure 2).

Physical Testing

As in any data network, testing breaks down into two parts: verification of the physical layer and interpretation of the received data. Data analysis will not be discussed here because it has little to do with the fiber itself. Rather, it is a system problem that has more to do with the amount of traffic, an asynchronous transport mechanism (ATM) protocol problem, or some other high-level incompatibility. These kinds of issues are best solved using protocol analyzers, network management tools, and traffic simulators.

The most straightforward physical-layer tests are for continuity and loss. Optical power is measured with an average-reading meter that uses a silicon, germanium, or InGaAs sensor, depending on wavelength. Because the reading will be sensitive to data duty cycle, measurements usually are made with a continuous wave (CW) or a squarewave source. When the actual data source is measured, the reading only will be repeatable if the data pattern is constant.

Tests include measurement of not just the transmitted power, but also of return loss or reflected power as low as -50 dBm or less. At the opposite extreme, community antenna television (CATV) operates with relatively high power levels up to +20 dBm to ensure a good signal-to-noise ratio (SNR). These two test examples demonstrate the need for power meters and sources with wide dynamic ranges.

A source and a meter often are combined in a convenient, hand-held tester. Generally, two testers are required, one for each end of the fiber being tested. Some manufacturers provide data download from the remote slave unit to the master unit. In this way, a single operator can set up the test and read the results from one end. It is important to run power tests from both ends of the fiber because the location of a marginal splice or connector can make a difference.

Work done by Corning demonstrates just how difficult it is to interpret test results accurately. Although 0.5 dB is an acceptable loss for a MM fiber connector, the actual loss experienced in a system depends on the transmission modes being propagated. Long fibers naturally tend to attenuate higher-order modes, which means that after a few kilometers, most of the power is concentrated in lower-order modes near the center of the fiber. This condition is termed equilibrium modal distribution (EMD). In contrast, a fully filled fiber exhibits high- and low-order modes, and all modes have equal power.

Under EMD conditions, a fiber-optic system can be up to 15 dB more efficient than the worst-case design figures would suggest. Fiber loss is about 1 dB/km less for the EMD case, and connector loss will be a few tenths of a decibel less.4

Another type of test tool, the optical time domain reflectometer (OTDR), determines loss vs distance. A short pulse of light is sent down the fiber, and the reflected light caused by the fiber’s backscattering properties is measured. Backscatter together with absorption are the primary loss mechanisms in fibers.

Because the backscatter coefficient varies according to the particular cable being used, OTDR displays can incorrectly show gain instead of loss if a cable with a different coefficient has been encountered. Other limitations of OTDRs are their distance resolution, the smallest distance that can be resolved, and the recovery distance required before the ambient power level is displayed again after a large signal change.

A good OTDR with modular sources to match different operating frequencies and types of cable also is relatively expensive. Regardless of these drawbacks, the OTDR is the instrument of choice for finding the location of fiber breaks or bad connectors and splices with high losses.


It is possible for a fiber to pass all continuity and power-related tests and still appear to cause data errors. If the bandwidth of the transmitter or receiver is too low, single data pulses may not be correctly converted to an optical or electrical signal, respectively.

One way to check bandwidth is to attach the receive end of a fiber to an optical-to-electrical (OE) adaptor. This device allows you to view the data pattern on a digital oscilloscope as a traditional eye pattern (Figure 3).

On some oscilloscopes with mask testing capabilities, you can create a test mask that is larger than the reference mask by a selectable amount. The new mask is used to determine the performance margin of the transmit/receive link above the guaranteed minimum level. If you want to see what the optical signal actually looks like, you need an OE with a much higher bandwidth than the bit rate, for example 5:1 although only 2:1 or 3:1 may be attainable at very high data rates.

The output from the receiver’s photodiode sensor is a bandlimited analog signal that, in turn, is converted to a digital bit stream. Bandlimiting occurs because the bandwidth of a receiver is designed to be not much greater than 80% of the bit rate or 8 GHz for a 10-Gb/s system. OE converters have switchable filters built in that provide pulse response characteristics similar to those of the actual fiber receiver.

Turning on the filter allows you to view the signal as the data recovery circuitry immediately following the photodiode element would see it. If you want to determine the bandwidth of the fiber communications channel, turn off the filter. You need all the bandwidth you can get. Make sure the oscilloscope’s bandwidth doesn’t degrade your measurements.

You also may want to run a bit error rate test (BERT). The BER is a good overall measure of fiber performance, and specialized test instruments are available to make this measurement. In a recent article about optical receiver elements for OC-192 (10-Gb/s) networks, the authors referred to a BERT performed with a 10-Gb/s, 223-1 pseudorandom bit sequence. The article compared the performance of a PIN diode and an avalanche photodiode (APD) in a receiver with 1xE-10 BER.

The authors showed that, because of the APD’s lower inherent noise, the optical receiver can be made about 5 dB more sensitive while achieving the same BER. The APD devices provided “increased SNR, longer transmission distance, and wider dynamic range.” These improvements are particularly important at 10 Gb/s, “where transmission limitations (fiber dispersion, optical nonlinearities, and insertion loss) put stringent requirements on receiver performance and specifications.”5


  1. Hayes, J., et al, A Practical Guide to Testing Fiber Optic Components and Networks, Fotec, 1997, Chapter 4.
  2. Girard, A., “DWDM Spurs All-Optical Network,” EE-Evaluation Engineering, Oct. 1999, pp. 30-36.
  3. Weinstein, C., “Fiber Design Improves Long-Haul Performance,” Laser Focus World, May 1997.
  4. Hayes, J., et al, ibid.
  5. Rue, J. and Nessar, B., “High-Speed Avalanche Photodiode Optical Receivers,” Fiberoptic Product News, Nov. 1999, pp. 27-32.

Published by EE-Evaluation Engineering
All contents © 2000 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.

February 2000

Sponsored Recommendations


To join the conversation, and become an exclusive member of Electronic Design, create an account today!