DWDM Communications Rely on Basic Test Techniques

Even as DWDM techniques cram more signals onto optical fibers, many tests still rely on standard equipment and procedures.

Jon Titus, Editorial Director -- Test & Measurement World, 3/1/2000

Thomas Edison figured out how to simultaneously send several signals down a pair of telegraph wires. Whatever his techniques, he expanded the signal-carrying capacity of existing wires. Today’s fiber-optic communication systems are undergoing a similar expansion as communication-system designers and builders use wavelength-division multiplexing (WDM) to increase the capacity of existing systems. These systems use a single optical fiber to carry many wavelengths—each of which can operate at gigabit-per-second rates.

Because newer systems densely “pack” as many as 40, and in some cases up to 80, discrete wavelengths on one fiber, the technique now called dense wavelength-division multiplexing, or DWDM. Future systems may expand the capability of fiber-optic (FO) communication systems to more than 200 wavelengths. And each wavelength—typically called a channel—may have a bandwidth up to 10 GHz. At this bandwidth, a 40-channel system could transmit a 10,000-volume encyclopedia in 1 s.¹ Test engineers need to know how such DWDM systems work and what characteristics they must test for to ensure optimum operating conditions. Lost signals mean lost income—possibly millions of dollars per hour of downtime. Tests cover the optical characteristics of a system as well as a system’s ability to accurately transfer information. This article describes several common optical tests so you’ll have a feel for what’s involved. Most of these tests rely on standard optical test equipment.

Before you can understand what tests you need to make and the results you should expect from them, you need to take a closer look at how a DWDM system works and the components in such a system, as shown in Figure 1. Digital communication signals enter the system at the left.

A DWDM system can use a single communication protocol on all channels, or it may mix IP, ATM, and other protocols as needed. The system does not care how users encode the information it transmits, nor does the system care at what bit rate the transmissions take place. Some channels might use slower SONET OC-12 rates while others operate at OC-192 rates.

The information in Table 1 shows the signal classifications and their data rates in the SONET hierarchy. Of course, the transmitters, receivers, and intervening components must be capable of handling the needed data rates and protocols.

In a DWDM system, electronic digital inputs modulate individual solid-state lasers, called distributed-feedback lasers, each tuned to a different but well-defined and standardized wavelength. A multiplexer ( Fig. 2) combines the various wavelengths into one optical signal that it then couples to a single fiber. Multiplexers may use filters or an optical diffraction grating to combine multiple wavelengths into one optical signal. Figure 3 shows a commercial multiplexer that combines 32 wavelengths. Multiplexers and demultiplexers may also use phased arrays (phasars) and filters to separate and combine optical signals. The technology chosen depends on wavelength spacings and the number of channels in use.

Figure 1. A simple DWDM system combines optical-communication signals at many wavelengths and transmits them over a single fiber. An intermediate add-drop tap picks off one wavelength and inserts other information in its place.

Useful Wavelengths Come in a Narrow Band
At first you might imagine the lasers produce a broad range of wavelengths to create a “rainbow” of signals. Actually, all the wavelengths must exist in a narrow band, usually between about 1530 nm and 1565 nm. This wavelength band was not chosen randomly. It corresponds to wavelengths that undergo the least attenuation in optical fibers, and it also corresponds to the band of wavelengths amplified in erbium-doped fiber amplifiers (EDFAs).² Without these amplifiers, which optically “pump” erbium ions to amplify optical signals, using FO links to communicate over long-distances would prove difficult. A DWDM signal may pass through many EDFAs as it goes from the transmitter to the receiver in an FO communication system.

The FO path may also include optical components such as isolators, Bragg gratings, and circulators. Combining two circulators with a Bragg grating lets the system switch out one wavelength at an add-drop tap as shown in Figure 1. An intermediate FO system can use the tap to fill the now-unused l₃ wavelength with information it wishes to send to the receiver. The illustrated tap removes wavelength l₃, but that choice is arbitrary. Designers can remove and add wavelengths as needed along the entire course of a FO system.

Figure 2. An optical multiplexer acts like a diffraction grating to combine several wavelengths into a single optical signal. A demultiplexer works in reverse fashion, separating a signal into individual wavelengths.

Figure 3. A commercial multiplexer provides pigtail inputs for optical signals. This unit combines 32 discrete wavelengths to produce one optical signal for transmission through an FO system. (Courtesy of Radiant Research.)

At the receiving end of the FO system, a demultiplexer splits the optical signal into individual wavelengths and routes those signals to opto-electric converters that reconstruct the electronic digital signal. Like an optical multiplexer, an optical demultiplexer works like a prism or a diffraction grating. A demultiplexer, however, divides the multiplexed optical signal into individual wavelengths.

Narrow Spacing Packs in the Wavelengths
In today’s DWDM systems, wavelength spacing occurs at 0.8-nm, or 100-GHz, steps. Future systems will space wavelengths at 0.4 nm, or 50 GHz, to further increase the capacity of existing fibers. The ITU has set specifications that list the channel code and wavelength for each optical signal multiplexed into a fiber.³ Present ITU specs recommend 81 channels, starting at a center frequency of 196.10 THz (1528.77 nm) and decreasing in 80 steps of 50 GHz (~0.39 nm) each. The ITU uses as its starting point a 193.10-THz reference frequency that corresponds to a spectral line for krypton gas. The ITU specified frequency rather than wavelength because wavelength depends on a transmission medium, and thus it can vary. Frequency is a physical unit that will not vary as a signal goes from one medium into another, say from glass into air.

Now that you know the basics of how a DWDM system operates, what do you have to test for? Plenty. Thorough testing can require dozens of tests to check the performance of a system and its conformance to standards. The ITU provides testing guidelines in several documents. The tests run on a DWDM system fall into four main categories:

• tests at the transmission side of the system,
• tests of the optical components in the system,
• tests on the fiber-optic paths, and
• tests at the receiver side of the system.

This article focuses primarily on tests at the receiver and transmitter sides of the FO link. The majority of tests require standard optical test equipment such as an optical power meter, an optical spectrum analyzer (OSA), and an optical wavelength meter. The latter two instruments should measure wavelengths with an accuracy of greater than 0.01 nm. A few tests may require an optical time-domain reflectometer (OTDR), an opto-electric converter, a scope, or specialized equipment. Your suite of test instruments will also include optical switches, optical attenuators, and laser sources. (See “Footnotes” and “For Further Reading” for publications that describe individual DWDM tests in more detail.)

Start at the Transmitter
At the transmitting side of the system, you’re most concerned about the quality of the optical signal produced by the lasers. So you’ll measure the laser signal’s characteristics, which can include wavelength, wavelength deviation, output power, side-mode suppression, and chirp, to name a few.

Figure 4. Sixteen multiplexed optical signals exist on background noise from an EDFA. Note the channels are not spaced perfectly and the background exists at about –25 dB.

Most people think of the spectrum of a laser as exceptionally narrow—and it is when compared to other light sources. Unfortunately, the bandwidth of an OSA is not good enough to accurately represent the narrow signal produced by a laser. Thus, the DWDM signal shown as an OSA output in Figure 4 produces peaks that aren’t single wavelengths. Note, too, that the wavelengths are not evenly spaced. Manufacturers cannot produce practical lasers that exactly match the frequency (wavelength) spacings set by the ITU. Most laser manufacturers, however, do keep the wavelengths within 10% of the ITU’s specifications. Manufacturing variations account for the small wavelength variations shown in Figure 4. That signal comes from the output of an EDFA, and the amplifier reduces the signal-to-noise ratio to about 25 dB. The combined laser signals ride on top of amplified spontaneous emission from the EDFAs. The SNR for the laser signals alone is about 40 dB.

Perhaps the best way to start looking at a system is by measuring the characteristics of the transmitters. The center wavelength characterizes each laser and lets you determine whether it operates at one of the ITU-specified wavelengths. Use the peak amplitude value to identify the laser’s wavelength, which should correspond with its specified ITU wavelength, within the 10% allowance. You can make the wavelength and power measurements using an OSA. Although an OSA cannot completely resolve the laser signal, the peak value is easy to measure and should provide a consistent value from measurement to measurement. The laser diode should run in continuous-wave (CW) mode when you make this measurement.

Wavelengths Can’t Wander
Changes in a laser’s wavelength can cause problems in adjacent channels. You need to test to ensure that any wavelength variations are small compared to the spacing between adjacent channels. The system cannot have a laser signal “wandering” into an adjacent channel. Unfortunately, the ITU specifications do not specify where the tolerance in laser frequencies comes from—manufacturing variations, component variations, or other effects. You’ll have to use the manufacturer’s design specifications for the DWDM system you’re testing. Those specifications may be narrower than the 10% allowed by the ITU. If specs aren’t available, use the ITU’s 10% tolerance.

Although manufacturers make very stable laser modules, optical and electrical effects can slightly change a laser’s center wavelength. You can use a multiwavelength meter to observe any wavelength variations that occur under simulated operating conditions. Remember, too, that wavelength can drift over time and under different operating conditions.

In addition to a fundamental signal, lasers also generate side modes offset from the main peak. An OSA will let you observe the side modes so you can be sure they remain low enough to keep from interfering with signals in adjacent channels. Typical laser-diode modules keep side modes to about 45 dB or more below the peak output. Thus, almost all of the laser’s power appears in its main signal, and not offset at wavelengths that could interfere with neighboring channels.

At the transmitter, you’ll also need to measure power output, or launch power, from the laser. The output power a system needs depends on the type of communication protocol the FO system employs and on whether transmissions take place over long-haul or short-haul links. The SDH, SONET, Fibre Channel, and Gigabit Ethernet specs spell out acceptable power ranges.

An optical power meter can make the measurements, which can go as low as –40 dBm. As in all measurements, be sure that the instrument you plan to use has spectral characteristics that match your signal’s characteristics. And be sure that you use a realistic signal source to stimulate the laser so it produces the type of signal the FO system will actually carry.

Don’t Let Transmitters Chirp
If the transmitters in a DWDM directly modulate the lasers, you may need to measure chirp. Rapid changes in power levels that occur at bit rates of about 10 GHz can slightly alter the optical characteristics of solid-state lasers, causing the output to “chirp.” The chirp appears as a small amount of added noise and as slight, but abrupt, changes in wavelength that depend on the laser’s construction. Chirp may broaden the laser’s output, which results in chromatic dispersion, an ill effect that slightly spreads the arrival of the laser’s signal at the receiver. If the transmitters use separate modulators, chirp should not be a problem in the DWDM system. You’ll need specialized equipment to make chirp tests.⁴

You may choose to make eye-pattern, or eye-diagram, measurements on transmitted signals. Eye diagrams tell you whether or not the signal characteristics meet specific ITU specifications. You also can use eye-diagrams to measure a signal’s extinction ratio—the ratio between the laser’s power output for a logic zero and for a logic one.⁵ You want to ensure that the system’s receivers can tell the difference between the two logic levels. Typically extinction ratios range from about 8.2 to greater than 10. These eye-diagram tests require carefully selected opto-electric converters and a wide-bandwidth scope.

Receivers Have Split Ends
At the receiving end of a DWDM system, the demultiplexer gets the most scrutiny because it must faithfully split the multiple-wavelength signal into individual wavelengths. You must make measurements of insertion loss, crosstalk, and polarization effects for the demultiplexer. You’ll need to make other measurements, too, but these illustrate some of the measurements typically made on DWDM systems.

Insertion loss simply describes how much the demultiplexer attenuates a signal at a given wavelength. You measure the power at a specific wavelength prior to its reaching the demultiplexer and then repeat the measurement after the signal leaves the demultiplexer. The difference, in decibels, tells you how much power gets lost in the demultiplexer. If too much power is lost, perhaps as evidenced by high bit-error rates, you will have to increase the power at the transmitter or through the system to ensure that enough power reaches the opto-electric receivers.

Sometimes when power reaches a receiver, it comes simultaneously from adjacent channels. The blending of these signals causes crosstalk, which means a receiver gets information from its corresponding optical source and also from one or both of the channels on either sides. In most cases, crosstalk includes only small amounts of power from adjacent channels, but even small amounts can affect receiver performance. Crosstalk can arise from laser wavelengths that drift into an adjacent channel or a demultiplexer that does not effectively divide the optical channels. You can measure crosstalk using the results from an OSA.

In a DWDM system, you compute crosstalk of l_x by first measuring the maximum insertion loss for the l_y signal. It doesn’t matter whether or not the maximum loss occurs at the band edge closest to l_x; you simply want the maximum value. Then you locate the band-edge wavelength of l_x on the side of the signal closest to the l_y channel. The points at which the l_x signal crosses the –1-dB line locates the band-edge wavelengths (Fig. 5a). Next, you measure at the l_x band edge the difference between the maximum power input to the multiplexer for l_y and the power output from the multiplexer for the same wavelength (Fig. 5b). The difference between the maximum insertion loss for l_y at the band edge and the power of l_y at the adjacent signal’s band edge gives you the crosstalk value, expressed in units of decibels.

Figure 5. (a) Bandwidth measurements occur at the –1 dB and –20 dB points for a laser signal. (b) You can use those measurements to help you compute crosstalk between adjacent channels, lx and ly.

You may find other ways of determining crosstalk, but this technique provides the most conservative value. Typically, crosstalk in channels spaced at 100 nm should be less than –20 dB, and it can reach as low as –45 dB. The larger the negative number, the less crosstalk exists. The crosstalk specification for a DWDM system comes from the system designers, who know how much crosstalk it can accept.

If the system designers can decreases crosstalk, they can place channels closer together. But the close proximity comes at a real cost. The equipment that operates at close wavelength spacings costs more than equipment that works with wider-spaced wavelengths.

Polarization Affects DWDM Performance
The polarization of the laser signal also plays a role in the performance of a DWDM system. Polarization can affect signal loss and the accurate reception of signals. You can measure polarization effects using a polarized broadband source, a polarization controller, and an OSA (Fig. 6). The controller lets you adjust the polarization of the light reaching the device undergoing testing. The test setup also includes a tunable laser and a wavelength meter that help calibrate the OSA during use. An OSA alone may not have enough resolution to precisely measure band edges in polarization testing. By including a tunable laser and a wavelength meter in your test setup, you can accurately measure the locations of band edges. The OSA provides fast measurements to locate areas of interest and then the wavelength meter measures them accurately. The optical switches in the test setup make it easy to configure tests without making changes by hand.

Figure 6. Polarization-dependency tests of a demultiplexer rely on standard optical test equipment. Two optical switches make it easy to change the test configuration without disturbing the equipment.

The power loss in a demultiplexer shows a dependence on the polarization of the incoming signal (Fig. 7). As the polarization of the light reaching the demultiplexer changes, so does the loss. The power loss between the maximum and minimum value forms a band called polarization dependent loss (PDL). This loss forms a small part of the demultiplexer’s insertion loss. Because system designers cannot control the polarization in a DWDM system, you must take into account the maximum possible loss due to polarization. You can expect a PDL that may range from a high of –0.5 dB to as little as –0.05 dB, although designers aim for the latter value; values typically reach as low as –0.1 dB. Theoretically, you should add this loss to the maximum insertion loss, but in most cases, the value doesn’t change the overall response of the system. In some optical systems, however, PDL can reach as high as several decibels.

Figure 7. The loss in a demultiplexer shows a dependency on the polarization of the input signal. The difference between the minimum and maximum values yields the polarization-dependent loss value.

The measured center wavelength of the arriving signal will also vary slightly, depending on the signal’s polarization. The actual center wavelength remains the same, but the polarization will affect the signal as seen at the receiver by shifting the signal (Fig. 5a) slightly to the left or right, or it will narrow or broaden the signal slightly. The polarization of the signal causes filters in the demultiplexer to slightly change the optical signal.

By varying the polarization of the test signal and measuring the center wavelength, you obtain data that lets you determine the maximum variation of the center wavelength. This variation goes by the long name “polarization dependence of the center wavelength,” or simply PDC. Test engineers must measure PDC so designers can figure its value into their development of a DWDM system because PDC will affect the performance of demultiplexers.

The variation of the center wavelength in effect reduces the 1-dB bandwidth of the demultiplexer. To avoid excessive loss of signal, the center wavelength of the transmitter and its wavelength drift must occur within the reduced bandwidth.⁶ 

These measurements should give you a better idea of what testing a DWDM system and its components involves. For the most part, tests use basic test equipment to make fairly standard measurements. By combining the tests and the test results, you and the design engineers get a better understanding of how a DWDM system will operate and where to look for possible problems. You can perform many other tests, such as bit-error tests that check on the quality of the information transferred from the receivers to the transmitters. And you can perform other optical tests that tell you about the performance and “health” of subsystems and components. T&MW

FOOTNOTES

1. Kartalopoulos, Stamatios V., Introduction to DWDM Technology: Data in a Rainbow, IEEE Press, Piscataway, NJ. 2000.

2. Hentschel, Christian, “Characterize DWDM Optical Amplifiers,” Test & Measurement World, November 1998, pp. 23–28.

3. ITU-T Recommendations: G.691, “Optical interfaces for single-channel SDH systems with optical amplifiers;” G.692, “Optical interfaces for multichannel systems with optical amplifiers;” and G.957, “Optical interfaces for equipments and systems relating to the synchronous digital heirarchy.”

4. “Performance and Conformance Testing of Complex Photonic Networks,” Tektronix, Beaverton, OR. (In preparation.)

5. LeCheminant, Greg, “Key Characteristics Govern OE-Converter Choices ,” Test & Measurement World, March 1999, pp. 11–14.

6. “The DWDM Components Test Guide,” Agilent Technologies, Palo Alto, CA. 1996.

FOR FURTHER READING

“Introduction to WDM Testing,” EXFO, Vanier, QC, Canada. 1997.

Strassberg, Dan, "200 THz: testing's tough new frontier," EDN, July 6, 2000. www.ednmag.com/ednmag/reg/2000/07062000/14cs.htm

You can contact Jon Titus at jontitus@cahners.com.