One of the biggest myths surrounding polarization mode dispersion (PMD) is that it is only an issue with old fibers installed more than about 10 years ago. However, experience shows that there can be problems with cables that were installed during the optical boom when the established, reputable fiber manufacturers were unable to meet the demand for quality fiber. Poor-quality fiber that would otherwise have been scrapped found its way onto the market; fiber was also procured from other sources around the world whose fiber manufacturing processes and procedures were not as well controlled as those of the major vendors. In fact, the record for the worst fiber this writer has ever tested comes from this era: Can anyone beat a figure of 27 psec/√km for the PMD coefficient of an installed fiber link?
In the context of increasingly higher-speed services, PMD is the factor most likely to limit the maximum data rate that a fiber will support. Unfortunately, it is not uncommon for different fibers in the same cable to have completely different PMD characteristics. This means that operators now need an extra dimension to their record keeping to make the most efficient use of their installed fiber plant. In addition to the cable-based records of how many fibers they have, where they go, and what the losses are, there is now a need for fiber-by-fiber records of what data rate each fiber can support. This allows poorer fibers to be allocated to lower-data-rate services (up to OC-48, SDH-16, or OTN1); better fibers can be used for 10 Gbits/sec; and the best performing fibers can be reserved for future operation at 40 Gbits/sec or maybe even higher data rates.
Of course, to populate these fiber records it will be necessary to measure the PMD of every fiber in every cable. Such PMD testing also is often a contractual requirement for dark fiber deals, as part of a comprehensive fiber characterization exercise (see “Characterization is Key to Exploiting Fiber Potential,” Lightwave, April 2006, page 22). So if PMD testing is required, how do we do it?
The IEC 61280-4-4 standard, published earlier this year, defines six main test methods for testing of installed fiber-optic links. Since all the techniques have their particular strengths and application areas, there is no preferred “reference test method.” Some of these six methods, however, have only been implemented in a laboratory environment-often using large, expensive, and delicate equipment.
With practicality in mind, let’s look at the three PMD test methods that have been implemented by a number of test equipment manufacturers in portable, robust field test platforms. Often these platforms are modular and can carry out many different optical tests to provide complete fiber characterization. Some of the platforms also support application-layer testing. Although measuring the PMD of every fiber can be a significant investment in time and money, choosing the right test platform for your application can make the task much more affordable.
Method A: Fixed analyzer. Instruments based on the fixed analyzer method launch polarized light into one end of the system and investigate how the received power through a fixed polarization analyzer varies as a function of wavelength. The higher the PMD, the greater the variation in received power as a function of wavelength will be. Figure 1 shows this variation for one implementation of the technique using a tunable laser as the light source; this fiber is 144 km long and has total PMD of 4.2 psec.
Another implementation of the technique uses a polarized broadband light source at one end of the system and a spectrum analyzer at the receive end. The spectrum analyzer makes two scans, one with and one without the fixed analyzer in the optical path. The resultant trace of the ratio of these two scans is then computed and a Fourier analysis is conducted to determine the frequency content of the trace to yield the PMD.
This second version of the fixed-analyzer technique has been implemented by JDSU in its T-BERD 8000 platform. In this case the broadband light source, covering the entire O-, E-, S-, C,- and L-bands, occupies a single module slot in one T-BERD unit at the launch end; a single module slot in the receive unit houses the optical spectrum analyzer, with an integral fixed analyzer that is automatically switched in and out of the optical path. This setup has the benefit that the same hardware can also be used to measure spectral attenuation (the loss of the fiber as a function of wavelength over the entire wavelength range from 1,250 to 1,650 nm), as well as carrying out measurements on CWDM and DWDM systems.
When Sunrise Telecom acquired Luciol, the company obtained PMD expertise that has now been packaged into modules for the STT test platform. This set implements the fixed analyzer method of PMD testing using a compact standalone C- and L-band broadband light source that is also used as a light source for chromatic dispersion measurements. The dispersion analyzer occupies a slice in the main STT platform.
An alternative implementation of the fixed analyzer technique appears in the Chromos 11 and 12 from PE.fiberoptics (formerly PerkinElmer). A tunable laser is used as the light source to provide dynamic range of 60 dB. The receive unit monitors the power levels through the fixed analyzer for each wavelength and then communicates with the remote laser source via another fiber (or an IP link in the Chromos 12) to change to the next wavelength. Since the wavelength scan is built up over some time, it is important that the fiber does not move during the measurement.
Method B: Stokes parameter evaluation. Using the Stokes parameter approach, the differential group delay (DGD) at a particular wavelength may be measured directly, rather than just the average value of DGD (i.e., the PMD) that is determined by the other test methods. The DGD is what directly affects the system performance, so it can be important to know what the DGD is for a particular channel of a DWDM system, for example. This method also allows you to determine how much the DGD varies as a function of wavelength for any wavelength range. This is known as second-order PMD and gives rise to a form of “polarization-dependent chromatic dispersion.”
The measurement is carried out by launching laser light into the fiber in three different polarization states and using a polarization analyzer to investigate the polarization state of the received light for each input state. The relationship between these may then be computed using some complex math called “Jones Matrix Eigenanalysis” to yield the DGD.
Agilent has managed to implement this complex test method in its Modular Network Test (MNT) platform. A standalone tunable laser with integrated polarization controller is used at the launch end, and the polarization analyzer is housed in the N3909A module in the MNT. As the tunable laser carries out a wavelength sweep, it simultaneously changes the polarization of the launch at very high frequency. The receive analyzer monitors what goes on and then, using the fiber under test as a communications link, the transmitter unit transfers information regarding the wavelength scan just carried out to the receive unit to allow it to compute the DGD as a function of wavelength over the tuning range of the laser. Each wavelength scan takes just a few seconds and the measurement is usually complete after two or three scans.
A plot is produced that shows how DGD varies as a function of wavelength. The average value of the DGD is the PMD of the link.
Method C: Interferometry. The traditional interferometric technique has historically been the most common approach for the measurement of PMD on installed fiber-optic links. The measurement uses a broadband LED as the light source; the output from the LED is linearly polarized before being launched into the fiber under test. The receive unit has another polarizer and then the light goes into a Michelson interferometer. A scan is carried out as the moving mirror of the interferometer passes through the reference position. The PMD of the fiber under test causes interference to occur at mirror positions away from the reference position. This is shown on the “interferogram,” a plot of the envelope of the interference pattern as a function of mirror position. The displacement of the mirror position is usually expressed in picoseconds as the resultant time delay introduced into the light path by that amount of mirror displacement.
For a typical fiber link this interferogram will have a characteristic Gaussian distribution. The PMD is calculated from the width of this Gaussian shape (see Figure 2). Examples of systems that use this technique are the Anritsu NetTest CMA 5000 and the PE.fiberoptics ONA600.
In 2003, EXFO proposed an enhanced interferometric test method that became known as GINTY, from Generalized INTerferometrY. At the same time the nomenclature of Traditional INTerferometrY (TINTY) was adopted to distinguish the two methods.
The GINTY system adds polarization scramblers before and after the fiber under test as well as a more complex detection system that puts a polarization beam splitter onto the output of the Michelson interferometer so that two interferograms are generated using orthogonal polarizations. These two interferograms are then processed to eliminate the autocorrelation peak.
GINTY also uses a different analysis that removes the assumption that the interferogram follows a Gaussian distribution. EXFO uses this technique in the FTB400 multipurpose test platform. The GINTY receive unit, the FTB5500B, is housed in a three-slot module that makes it necessary to use the seven- or eight-slot back end to the FTB400. A separate standalone light source is used for both PMD and chromatic dispersion measurements; it contains a broadband, polarized LED with modulated output.
With this many options, how do you choose the right instrument for your application-and how do you ensure that you get valid test results? Several factors can influence both the choice of system and how it should be used: configuration, measurement range, fiber movement, number of tests, and portability.
Let’s look at these factors individually. The vast majority of PMD testing is carried out on installed links that may be anything up to 100 miles (160 km) long, although most are usually less than half that distance. All of the aforementioned systems have enough dynamic range (typically 47 dB or more) to measure typical 160-km links with modest amounts of PMD. Note that the manufacturer’s data sheets often quote best-case dynamic range figures-with most systems as the amount of PMD increases, the dynamic range decreases. The dynamic range for 10 psec of PMD may be 8 dB less than for 1 psec.
There are upper and lower bounds to the amount of PMD that can be measured using some of the techniques, although these typically lie outside the range of PMD figures that are likely to be encountered on installed fiber links. In any case, for practical measurements on installed links, if the PMD is less than 100 fsec then it will be a “Pass” and the actual value is not usually important. Similarly, if the PMD is greater than 50 psec then it will be a “Fail” and again the actual value doesn’t really matter.
The interferometric technique is not adversely affected by movement of the fiber during a measurement, but other techniques may be affected. However, the very fast acquisition times of the Agilent MNT and the spectrum analyzer implementation of the fixed-analyzer method by JDSU and Sunrise Telecom limit potential problems.
In the commercial world of fiber characterization, time is money, so there is pressure to carry out as few measurements as possible. However, because some aspects of PMD are dynamic, some people think that it is necessary to carry out lots of measurements at different times of day and night, with different polarization states and different wavelength ranges. While all of these measurements enhance the statistical base, it is important to remember that the parameter that actually varies is the DGD. All of these test methods are already producing an averaged value of the DGD to create the PMD figure. On a good installation the fiber should be isolated from external stresses and strains; that is the job of the cable. Therefore, there should be no variation in PMD due to external factors. It is only the intrinsic PMD that is being measured.
So if you have the time, it is not a bad thing to carry out several measurements. If the measured value is close to a threshold value, then it is recommended to take more measurements to determine the distribution of values. Another approach that can be used with the interferometric technique is to use polarization mode scrambling during the measurement.
Another important factor for the user of the test equipment is portability: The test platforms equipped for fiber characterization range from about 14 lb (5.5 kg) up to 33 lb (15 kg). This can make a big difference if you have to carry the kit any distance!
There is now more choice than ever in equipment for testing PMD. Choice is good for the users of test equipment because it encourages competition and keeps the prices down. For real-world PMD testing of installed fiber links there are some very attractive, portable, cost-effective systems available. Well-informed purchasers who can see through the salesperson’s hype can get the right system that they need to do the job, without paying over the odds.
Richard Ednay is technical director of Optical Technology Training (OTT; www.ott.co.uk). He has more than 20 years of experience in fiber optics.