Eugene Rudkevich and Feenix Y. Pan
The use of polarization-dependent loss (PDL) calibration reference devices can improve the integrity of PDL measurements during component manufacture.
The optical networking industry is maturing and acquiring many similarities to other high-tech industries such as semiconductors, electronics, and computers. One feature that it increasingly shares with others is the need for accurate, repeatable, reproducible, and traceable parameter measurements in the physical layer. These requirements are partially accomplished by the use of measurement reference devices.
These devices, also known as artifacts and/or standards, are built to produce a specific measurement parameter with a specific value. They are then used to verify the integrity of the measurement systems used in production and R&D.
Many types of measurements are made on fiberoptic components, modules, and systems dealing with wavelength, optical power, and polarization. In this article, we will focus on polarization-dependent loss (PDL). Polarization-dependent loss is defined as "The absolute value or the relative difference between an optical component's maximum and minimum transmission loss, given all possible input states of polarization."1
A component with PDL acts like a partial polarizer with a low extinction ratio (see Fig. 1). Polarization-dependent loss is present to some extent in nearly every fiberoptic component. Some of the components in which PDL is large enough to be an issue are fused-fiber devices, liquid-crystal devices, planar waveguide devices, thin-film filters, and isolators. It is a detrimental parameter that can cause an increase in the bit-error rate when it interacts with polarization-mode dispersion in a fiberoptic link.2 Because of this, component and system manufacturers require accurate characterization of PDL to calculate power budgets and to remain within the maximum PDL specifications.
Two of the most common PDL measurement methods are the all-states method—also known as the polarization scrambling method—and the Mueller matrix method—also known as the four-states method (see Figs. 2 and 3). Standard procedures and test setups for these methods are also documented by the Telecommunications Industry Association and the International Telecommunications Union (ITU). Although both of these methods can produce accurate measurements, significant errors are possible due to improper measurement procedure and the components making up the measurement system.
The all-states method measures the PDL of a component by launching a large number of different polarization states of light into a device under test (DUT) and then measuring the minimum and maximum transmission at the detector. When using this method, any source of PDL in the measurement path other than the DUT will contribute to the measurement uncertainty. This includes the polarization scrambler itself, any connections exhibiting PDL, and the detector used. This residual PDL cannot be subtracted from the overall PDL value because it is a vector quantity.
A second source of error is the instability of the light source. Any fluctuation of the light-source power during the measurement procedure will directly contribute to the measurement error because it will be interpreted by the measurement algorithm as additional PDL.
The final source of error is in the polarization scrambler. To make accurate measurements, the polarization scrambler must cover all areas of the Poincaré sphere uniformly and with the same frequency. Any deviation from this will result in uneven coverage of all the polarization states and will contribute to measurement uncertainty. User error can also be introduced when the polarization scrambling time is insufficient to cover an adequate quantity of the polarization states on the Poincaré sphere.
The Mueller matrix method differs from the all-states method in that four well-defined orthogonal polarization states of light are sequentially launched into the DUT, the optical powers of which are then measured. A matrix calculation is used to calculate the PDL. In addition, to reduce the effects of power fluctuations of the light source, a reference detector is often used.
This method is also susceptible to measurement inaccuracy due to improper procedure and the properties of the components making up the measurement system. Nonorthogonal polarization states generated by the polarization-state adjuster will contribute to measurement error. This condition can be caused by improper calibration of the polarization-state generator. Furthermore, any drift in the component PDL in the signal or reference paths during the time between the reference measurement and the DUT measurement will affect measurement accuracy as a result of a change in the power readings at the detector. Finally, any movement in fibers during the measurement or a change in ambient temperature will decrease measurement accuracy.
One way to ensure the integrity of a PDL measurement system is to test it with a device with a known value of PDL. Although any type of device with previously measured PDL can be used to perform this task (a coupler, for example), maximum value would be derived from this procedure by using a device that has been specifically designed for this application. Devices that produce a fixed and well-characterized PDL are currently available. A good PDL reference must possess the following characteristics:
- Flat PDL over wavelength. This is required so that measurements made at one wavelength will correspond well with measurements at other wavelengths.
- Good thermal stability. The PDL must remain stable over temperature so that the same PDL measurement will be made regardless of temperature fluctuations during the measurement.
- Long-term stability. The PDL must remain stable over a period of months or years.
- Well-characterized performance. All the parameters must be measured to give the end user confidence in the overall flatness and stability.
A PDL reference device can be used in several ways to improve measurement integrity:
- Calibrating the PDL measurement instrument or system. For example, if an instrument consistently measures a 0.15-dB PDL reference as 0.16 dB, then for future measurements 0.01 dB should be subtracted from the instrument reading. This procedure works best when the artifact PDL value is close to the PDL value of the DUT.
- Verifying the accuracy of a PDL measurement. If there is some uncertainty about a PDL measurement, then a calibration artifact can be measured to verify the proper operation of the instrument or system.
- Calibrating the temperature or wavelength dependence of a PDL measurement system. If a PDL reference has a flat and well-characterized wavelength and temperature dependence, then this device can be used to test the response of the PDL measurement system to variations in wavelength and temperature dependence.
- Comparing multiple measurement stations. In a larger production facility, station-to-station measurement uncertainty can be reduced by making PDL measurements at all stations on the same artifact.
- Comparing different measurement methods. In a production or R&D environment where more than one method is used for PDL measurements, it is essential to verify that both methods produce similar results or at least to document the differences.
- Comparing measurements at different sites. For example, a component supplier measures the PDL of a component in the final inspection, and then a component user measures the PDL as part of the acceptance test. To verify that both the supplier and the user are measuring the same PDL value, the same PDL reference can be measured by both the supplier and the user.
The following example illustrates why a PDL calibration reference would be used to reduce the rejection rate of manufactured components in a production environment. A PDL measurement system with a possible measurement error of ±0.03 dB measures 0.13-dB PDL on a sample component. If this component has a PDL specification of 0.15 dB, then there is no certainty that the component is within specification because it may in reality exhibit a PDL as high as 0.16 dB.
Under strict quality-control practices, this component would have to be rejected. However, if the possible PDL measurement error of a system is improved to ±0.01 dB, and the same component is measured at 0.13 dB again, this component would pass because the maximum PDL it could have is now 0.14 dB. Thus, improving the accuracy of the PDL measurement can improve production yields.
As with other measurable fiberoptic parameters, traceability of measurements is highly desirable. That is, a standards lab must be capable of producing an accurate PDL measurement that can be verified and well-accepted by the measurement community.
Established procedures in certifying the capabilities of secondary labs by the primary standards lab—which requires providing traceability—must be in place. According to the US National Institute of Standards and Technology, a Measurement Assurance Program (MAP) will soon cover traceability of PDL at selected ITU grid wavelengths in the 1530- to 1560-nm band. Unlike the Standard Reference Material program under which the polarization-mode dispersion and wavelength references operate, the MAP program will make a selection of several certified measurement artifacts available for loan rather than for sale.
New higher standards of quality are becoming a requirement for the maturing fiberoptics telecommunication industry. Polarization-dependent loss is an increasingly common parameter that is specified and measured. The use of PDL calibration reference devices is an important approach for improving the integrity of PDL measurements.
The authors would like to thank Rex Craig of the National Institute of Standards and Technology, Boulder, CO for useful discussions.
- Agilent Product Note 8509-1.
- E. Lichtman, J. Lightwave Technol. 13, 906 (1995).
Eugene Rudkevich is chief technical officer of Taliescent, 10088 E. Paseo San Rosendo, Tucson, AZ 85747; and Feenix Y. Pan is a Ph.D. candidate at the Optical Sciences Center of the University of Arizona, Tucson, AZ 85721. Eugene Rudkevich can be contacted at firstname.lastname@example.org.