Measuring loss from polarization sensitivity

Although the two common methods of measuring loss work best in different situations, they should produce similar results.

Gunnar Stolze

All passive optical components exhibit polarization sensitivity in the transmission of lightwave signals. As more optical components are deployed, their polarization-sensitive transmission must be taken into account in the power budget of fiberoptic-network links. Consequently, the measurement of this sensitivity, called polarization-dependent loss (PDL), has attracted enormous attention from component manufacturers.

FIGURE 1. With the polarization scanning method of measuring PDL, measurement uncertainty depends on the measurement time.

Because manufacturers now aim for the lowest polarization sensitivity when designing passive optical components, test and measurement of PDL has become a critical aspect in manufacturing. Measurement accuracy is not the only criteria requiring examination: engineers also must account for the huge variety of passive optical components including narrow or broadband filters. Combined with the demand for high manufacturing throughput and yield, this has created a need for precise but fast measurement methods.

Measuring polarization-dependent loss
Polarization-dependent loss is a measure of the peak-to-peak difference in power transmission of an optical component or system across all states of polarization. It is defined as the ratio of maximum-to-minimum transmission through the device, as shown in the following equation:

null

The polarization sensitivity of an optical component primarily results from angled optical interfaces, oblique reflection, dichroism, and fiber bendings. For this reason, a variety of components, including couplers, WDM components, photodetectors, and isolators, can have a high level of PDL.

When determining insertion loss and PDL of such devices, manufacturers typically rely on two methods. The most common technique, polarization scanning, is suitable for measurements at single, specific wavelengths. The other option, the Mueller method, works well when the component's PDL is required over a broad wavelength range.

In polarization scanning, the device under test (DUT) is exposed to a wide range of polarization states. The minimum and maximum transmissions are then measured with a power meter, and the resulting data is used to calculate the PDL.

FIGURE 2. Although they work best in different situations, polarization scanning and the Mueller method should yield similar measurements, as illustrated by data for a grating-based WDM filter.

With a typical PDL-measurement setup of this type, the source produces nearly fully-polarized light, while a fiber-loop-based polarization controller generates pseudo-random polarization states (see Fig. 1). Movement of the loops produces variations in the birefringence of the fiber, which in turn generates varying polarization.

Since it is impossible to expose the DUT to all states of polarization, end users generate the number of polarization states at a rate that aligns with the averaging time of the power meter. Called the polarization scan rate, this parameter indicates how fast the polarization of the lightwave signal changes. A fast scan rate generates more states of polarization in a defined time interval, which can decrease the duration of a measurement. Too fast of a scan, however, can invalidate results because the power meter will take an average reading based on more states of polarization. Ultimately, it could average out the impact of a maximum or minimum transmission.

Test engineers also must be aware of the impact averaging time may have on other parameters, specifically the signal-to-noise ratio. This parameter changes proportionally with the square root of the averaging time. Even the errors in maximum and minimum transmission will directly influence the PDL value. The longer a polarization scan takes, the smaller the uncertainty of the PDL measurement becomes as the DUT is exposed to more states.¹

The second common approach to determining PDL is the Mueller method, which calculates the loss by exposing the DUT to four well-known states of polarization: linear horizontal polarized, linear vertical polarized, linear + 45°, and right-hand circular.^{2, 3} The technique involves measuring the optical power at each state for calibration purposes, then applying the four polarization states to the DUT, and measuring transmitted optical power. Calibrat-ing these results against the reference measurements will eliminate the impact of power variations due to the measurement setup.

The Mueller process involves determining a 4 x 4 matrix (M_DUT) of the DUT, which describes its polarization and power-transmission properties. The relation between an input Stokes vector and output Stokes vector of a DUT can be written as S_out) = M_DUT) x S_in) . The four first-row coefficients of the matrix describe the power transmission of a device, which is sufficient to obtain the PDL. With the calibrated DUT power-transmission measurements, a system of linear equations can be solved to determine the desired coefficients of the Mueller matrix and derive the maximum and minimum transmissions used to calculate the PDL.

With the Mueller technique, the four polarization states are synthesized by a polarization controller, which usually consists of a quarter- and half-wave retarder. The retardation plates of the polarization controller transform a linear input state into any other state of polarization, allowing the desired state of polarization to be set using specific rotation angles of the retardation plates.

Loss measurement solutions
Both polarization scanning and the Mueller method yield the insertion loss of a device, which facilitates the calculation of other DUT measurement parameters, including channel-center wavelength, bandwidth, or crosstalk. To ensure the highest measurement accuracy, though, the test setup must meet certain requirements. First, the tunable laser source must have a stable power output. Otherwise, output variation, which is not recognized by a PDL measurement, could be interpreted as polarization sensitivity of a device.

FIGURE 4. Choosing the process to measure PDL is not always simple, especially with passive components that transform one input signal containing many channels into a corresponding number of output signals containing only one channel, or vice versa.

The setup also must take into account the accuracy and repeatability of the wavelength. Wavelength accuracy determines the absolute location of the filter curve along the wavelength axis. Wavelength repeatability is especially important for the Mueller method, in which the filter curve is measured four times at different input-polarization states. Any deviation in wavelength between the four measurements will heavily influence the PDL results.

Photodetectors are among the components that exhibit PDL, so it is essential to use detectors with low polarization dependence. The PDL of different components combines in an uncontrolled manner, which means the PDL of the detector may influence the measurement. Spectral ripple of the power detectors can be another problem.

Lastly, attenuated-source spontaneous emission and low noise level of the power detectors also are requisites for high dynamic range, which is important when determining device characteristics.

All these requirements are important when measuring PDL across a range of wavelength points, a task that usually sets a tunable laser as the source. The measurement method used will depend on the DUT. Ultimately, both measurement techniques should yield similar results (see Fig. 2).

Polarization scanning, for instance, can only measure one wavelength at a time, so capturing the PDL of a device at many wavelength points can become time-consuming. If, however, the PDL is only required at certain points, the technique will work well (see Fig. 3).

In contrast, if several channels must be fully characterized for insertion loss and PDL, measurement time can decrease dramatically by using the Mueller method in conjunction with a continuous wavelength scan at each of the four well-defined polarization states.

The choice is not always simple, however, especially with passive components that transform one input signal containing many channels into a corresponding number of output signals containing only one channel, or vice versa. A typical example involves a multiplexer/demultiplexer, in which it is desirable to characterize all output channels with a single measurement to reduce measurement and device-handling times. Both the polarization scanning and Mueller method are capable of parallel multichannel testing (see Fig. 4).⁴

Incorporating testing in the process
When incorporating testing in the manufacturing process to control and manage production quality, both measurement accuracy and reliability become extremely critical. Both types of PDL testing techniques may find use at different stages in the production process.

Consider the production of a multiplexer based on arrayed waveguide technology. Prior to the assembly and packaging of the multiplexer, the arrayed waveguide chip is characterized channel by channel in a pass/fail test. The testing process investigates each channel for its insertion loss and PDL, as well as parameters such as bandwidth and center wavelength. To reduce test cycle times, the channels are not characterized to their full extent. Instead, measurements may take place only at specific points, such as the center wavelength and 3-dB bandwidth. The results must fall below certain limits to pass the test. For these conditions, the polarization scanning technique is a better choice than the Mueller method.

Comprehensive device testing, which only occurs during final inspection, should measure PDL over a broad spectral range to fully characterize each channel component. Here the Mueller method, combined with a continuous wavelength scan, provides an accurate and fast solution.

REFERENCES

1. Agilent Technologies, " PN 5965.
2. B. Nyman, OFC 1994 Tech. Dig., 230, ThK6.
3. Agilent Technologies, PN 5964.
4. Agilent Technologies, PN 5980.

GUNNAR STOLZE is an application integration engineer with Agilent Technologies, Optics Communication and Measurement Div., Herrenberger Str. 130 71034 Böblingen, Germany.