Polarization modes complicate FO communications

-March 01, 2001

As fiber-optic (FO) transmissions take place over longer distances and at higher data rates, subtle effects can impair communications. At data rates above 10 Gbps, the undesirable effect of light polarization makes its presence known. Light exists polarized in the horizontal or vertical plane (called the principal states of polarization) or as a vector sum of horizontal and vertical components. Problems with data transmissions occur because light polarized in each plane can travel at different speeds, which leads to pulse spreading.

Even though optical-fiber manufacturers do their utmost to produce perfect fibers, single-mode optical fibers made of silica exhibit birefringence. Birefringence in optical fibers results from slight geometry changes and small imperfections in the material, and it causes an effect called polarization mode dispersion (PMD).

When a laser “launches” a light pulse into a fiber, the pulse contains both horizontally and vertically polarized light. Due to PMD, a small time difference—usually a few picoseconds or less—exists between the arrival of the polarized components of a light pulse at the receiving end of a fiber (Figure 1). You cannot tell ahead of time which polarized component will arrive ahead of the other. If the difference in arrival times becomes too great, it can lead to bit errors. Unfortunately, designers of FO systems have few ways to counteract the effects of PMD; at best, they can only measure its effects and take them into account.

In most cases, PMD occurs randomly along a fiber due to random mode coupling, so the effects of PMD can change with distance, time, and wavelength. Random mode coupling simply means that as the polarization’s electric field moves through a fiber, it randomly couples into different birefringent sections.

You may wonder why FO communication systems don’t simply rely on light sources that provide only one principal state of polarization and use optical fibers and components that maintain the polarization at all times. Although that sounds like a simple solution, implementing such a system proves more difficult and more expensive than carefully characterizing a system that contains birefringent components. You can buy polarization-maintaining fibers, but the optical components that launch a signal into the fiber and every component in the transmission path must carefully maintain the polarization. That task proves almost impossible over long distances.

Instruments measure delays

Using commercial instruments, you can accurately measure the delay caused by PMD at a specific wavelength and time. The delay usually goes by the formal name differential group delay (DGD) and is designatedDT. The value of DGD most commonly specified is its mean value, & DT>.

You also will see another value associated with randomly mode-coupled optical fiber. The PMD coefficient, often called the PMD delay coefficient expresses a delay that accumulates as the square root of distance. A typical single-mode fiber might have a reported PMD coefficient of 0.05 ps/EFkm.

This fiber would have a calculated DGD of 0.05 ps over 1 km, or 0.50 ps over 100 km. But these calculated values may not represent accurately the DGD of the installed fiber. Due to the nature of random mode coupling, PMD depends on wavelength and environmental effects. Although calculations can give you an idea of DGD, actual measurements on a system provide the most reliable values, but even those measured values may change slightly over time and over a range of wavelengths. In effect, DGD represents a statistical value calculated from measurements.

TMW0103F4FIG1.gif (12231 bytes)
Figure 1. Polarization mode dispersion (PMD) causes a slight delay between horizontally and vertically polarized portions of a light pulse over long fiber-optic (FO) links.

The environment plays a role

Strange as it may seem, environment plays a role in PMD. Ambient conditions change the stress on a cable, and the stress alters the cable’s birefringence. Simply unwrapping the cable from a spool, installing it, and subjecting it to various twists and bends will change its DGD. For FO cable mounted on telephone poles, stresses from wind and temperature alter the optical path, resulting in a DGD value that changes with time. (Now you know why so much FO cable gets installed underground.)

As data rates increase above 10 Gbps, DGD can significantly alter digital signals. A rule of thumb states DGD must not exceed 10% of a bit time, or bit error rates will become unacceptable. At 10 Gbps, a signal has a 25-ps bit time, so for effective communications at this rate, DGD cannot exceed 2.5 ps. The graph in Figure 2 shows the relationship between maximum transmission distance, PMD coefficient, and transmission rate.

Because you can’t accurately predict DGD, you must treat it as only a statistical measure of delay in a fiber or component. Thus, you use it to help predict a level of system performance rather than guarantee a specific performance. Unlike many physical phenomena, DGD follows a Maxwellian distribution, which works to designers’ advantage. Using this type of distribution, researchers have calculated that, in most cases, the DGD will exceed three times the mean value for only about 21 minutes per year. (Ref. 1) It’s always possible that DGD may exceed your acceptable error rate. The best you can do, though, is calculate its probable effect on a communication system.

 TMW0103F4FIG2.gif (18849 bytes)
Figure 2. Plotted data lets you determine the maximum PMD a fiber link can tolerate at a given distance for a given data rate. The charted data assumes a –1-dB power penalty. Courtesy of Profile Optische Systeme.

Buy, don’t build

Before you worry about DGD exceeding a set amount, though, you need to know how to measure it so you can gauge a system’s performance. In all likelihood, you’ll buy instruments that directly measure polarization characteristics and calculate DGD and related statistical information. The short time differences and the narrow wavelengths would make it nearly impossible for you to build your own instruments. Several manufacturers supply equipment for lab use and portable instruments for field use. (See “Manufacturers of PMD instruments”)

Other sources (see “For more information”) provide the details of instrument operation and the math used to make DGD measurements, so there’s no need to duplicate that information here. I will, however, briefly describe three measurement techniques to give you a general idea of what’s involved. Most instrument suppliers use one or more of the following standard techniques in their equipment: wavelength scanning, interferometry, or the Jones-matrix method. Other methods include the modulation phase shift, Poincaré sphere, pulse delay, and baseband curve-fit techniques. (Ref. 2)

TMW0103F4FIG3.gif (7133 bytes)
Figure 3. The wavelength-scanning technique measures amplitude of polarized light received by the OSA at each frequency scanned.
TMW0103F4FIG4.gif (10387 bytes)
Figure 4. Information from the wavelength-scan measurements provides information about the distributions of maxima and minima in a range of wavelengths. This waveform shows 13 peaks and valleys. Mathematical analysis of the data produces a mean DGD value.

All of these methods use a light source that transmits polarized light down a fiber. They differ in how they control the polarization of the light, how they detect the polarization states, and how they measure the polarization “information” at a receiver. In general, here’s what the three basic techniques do:

The wavelength-scanning technique, also called the fixed-analyzer technique, uses a broadband light source, a polarizer, a second polarizer (called an analyzer), and an optical spectrum analyzer (OSA), as illustrated in Figure 3. The light source and polarizer provide polarized light to the fiber undergoing testing. The analyzer separates the polarized light, and the OSA records light intensity as a function of wavelength over the narrow range of wavelengths of interest.

TMW0103F4FIG5.gif (10739 bytes)
Figure 5. The interferometry technique uses a Michelson interferometer in place of an OSA. The acquired fringe patterns in the interferogram provide data that yields a mean DGD value for the fiber undergoing testing.
TMW0103F4FIG6.gif (8270 bytes)
Figure 6. Although the block diagram of an instrument that employs the Jones-matrix technique looks simple, the technique requires accurate computer control of many functions and mathematical analysis of matrix information.

Software analyzes the resulting data and counts the occurrence of peaks and valleys (Figure 4). Combining those counts with wavelength values produces a mean DGD for the system under test.

Some instruments can shift the amplitude vs. wavelength time-domain data into the frequency domain using Fourier analysis. By applying statistical techniques to the results, an instrument can provide a mean DGD value. Fourier analysis provides a way to filter out high-frequency signals induced by the local environment. Those signals could appear as unwanted peaks or valleys in the time-domain data.


TMW0103F4FIG7.gif (5031 bytes) 

Figure 7. An instrument that uses the Jones-matrix technique provides a plot of DGD vs. wavelength.

The interferometry technique directly measures time using a Michelson interferometer in place of the OSA in the wavelength-scanning technique ( Figure 5). The interferometer produces an “envelope” of fringe patterns (interferogram) that can undergo direct analysis to produce a DGD if there is no mode coupling. If mode coupling exists, as it usually does in long fibers, the instrument mathematically analyzes the best Gaussian fit to the interferogram to yield a mean DGD value.

The Jones matrix, or Jones-matrix-eigenanalysis (JME) method, relies on a computer-controlled instrument that includes polarization filters, and detectors to determine polarization states at a set of wavelengths (Figure 6). The math involved is extensive, and requires matrix manipulations to produce a wavelength vs. DGD graph (Figure 7 ).

The technique employs a tunable narrowband source—a laser—that can provide three known states of polarized light to the fiber undergoing testing. The computer calculates a Jones matrix by relating the input states to the output states at a series of wavelengths. (You can find out more about Jones matrices and their attributes at Eric Weisstein’s Treasure Trove of Physics.)

Manufacturers of PMD instruments
Agilent Technologies
Palo Alto, CA

Model 8509B Lightwave Polarization Analyzer (Jones matrix) Exfo
Vanier, QC, Canada
Model IQ-5500 and IQ-5500ER PMD Analyzers (Interferometer) GN Nettest
Utica, NY
Model PMD 400 PMD Measurement System  (Wavelength scan)
Model PMD 440 PMD Measurement System (Interferometer). Editor's Note 10/24/03: The company has changed names:www.nettest.com.
Peabody, MA
Model WIN-PMD (Interferometer)

Profile Optische Systeme
Karlsfeld, Germany
PAT 9000 B Polarization Measurement System (Jones matrix, Wavelength scan,  Poincaré sphere)

Santec USA
Hackensack, NJ
Model PMD-6000B PMD Test System (Interferometer) Tektronix
Beaverton, OR
Advantest Q7760 Optical Network Analyzer  (Polarization phase shift)

You’ll have to determine which technique or techniques best suits your measurement needs by consulting with vendors. No one type of measuring instrument or technique will always prove better than another in a given situation. T&MW


1. Chbat, Michel W., “Managing Polarization Mode Dispersion,” Photonics Spectra, Laurin Publishing, Pittsfield, MA, June 2000. p. 100.

2. Derickson, Dennis, ed., Fiber Optic Test and Measurement, Prentice Hall PTR, Upper Saddle River, NJ, 1998. This book provides excellent detailed descriptions of PMD and DGD measurement techniques.

For more information

“Basic Note: PMD and Polarization,” Profile Optische Systeme, Karlsfeld, Germany. www.profile-optsys.com.   Look in the Download section for “Basic Note BN-9000.”

Cyr, Normand, “Measurement of Ultralow PMD in Optical Fibers and Components,” Application Note, Exfo, Vanier, QC, Canada, September 1999. www.exfo.com. Go to the Support section and look under Technical Documents.

Jon Titus

has written real-time software and designed embedded systems and computer/instrument interfaces. He worked in electronics for 10 years and spent nine years at EDN magazine prior to joining T&MW in 1993. he has a BS from WPI, an MS from RPI, and a PhD from VPI.

Loading comments...

Write a Comment

To comment please Log In

Currently no items