Optical test needs tuning

Mike Minneman is worried about the optical communications industry. As president of optical component and equipment vendor dBm Optics (Lafayette, CO; www.dbmoptics.com), he sees a lot of uncertainty within communications equipment companies about the accuracy and repeatability of fiber-optic component tests. Fortunately, there are steps that these companies and their suppliers can take to ensure that what a vendor says it shipped, the customer can verify it got.

"In the early days of the fiber-optics industry, if a company could ship a part, it could sell the part regardless of its compliance to its specifications," says Minneman. "Now that the boom is over, however, it is still common to have noncompliant materials being shipped."

The trouble, Minneman believes, is that testing practices at both the vendor and customer typically produce data variations larger than the tolerances on the product specifications. As a result, vendors test products and send them to customers, and then the customers get a different result during incoming inspection tests and send the products back.

Paul Williams, a physicist in the Optoelectronics Division at the National Institute of Standards and Technology (NIST; www.nist.gov) in Boulder, CO, agrees with Minneman. "People have been willing to put out products without standards and measurements," he says, "and test procedures do lag behind the technology." But Williams also notes that optical testing is more difficult than testing within the electronics industry.

Part of the reason for that, says Williams, is the frequency of the electromagnetic waves involved. "In electrical testing, you can measure electric field strengths and phases for the signals involved, but in optics you can't." He notes that fundamental measurements such as signal power must depend on examining the signal's effect rather than on measuring the signal directly.

The lack of direct measurement makes it hard to be precise, explains Stefan Loeftler of Agilent Technologies' Photonic Measurement Division (Boeblingen, Germany). "The best accuracy available for power sensors still has a few percent variation in reading across the surface," says Loeftler, "so you have to get close to the fiber to use a homogeneous region of the sensor. But that risks interference effects. If you tilt the sensor to eliminate interference, you get polarization-dependent losses." All these factors combine to limit the achievable accuracy of measurements. Typical power measurement uncertainty runs about 1.5% with today's instruments, Loeftler says (Figure 1).

In another example, Williams points out that while the coefficient of thermal expansion for fiber material is small, at the wavelengths of optical communications even a 1° temperature change can alter the length of fibers by many wavelengths. As a result, the reflections from fiber ends form a standing wave pattern that can change significantly with temperature, affecting power readings. To see equivalent variations in RF measurements requires dimensional changes on the order of several centimeters.

Then, too, some parameters are inherently variable and need statistical measurements. Polarization mode dispersion, says Williams, is a bi-refringence effect that varies with the fiber's internal stresses. Make a measurement and then move the fiber, he says, and you'll get a different result.

Not everything in optics is that ambiguous, fortunately. "We can make highly accurate measurements in the time domain, for parameters such as frequency and wavelength," says Loeftler. He notes that instrumentation in this domain can reduce ambiguity to parts per million.

Still, the challenges in many aspects of optical testing in the communications industry remain significant and test engineers need to take special precautions to ensure accuracy and repeatability in their measurements. Williams and Loeftler recommend that test engineers take several actions:

• Analyze the test steps and make certain you understand the underlying manufacturing process being tested. This helps lay the foundation for implementing manufacturing process controls.
• Examine the test procedures for unintended sources of error. The test for insertion loss shown in Figure 2 seems like an appropriate setup, but it introduces new connections (C3/CX and CY/C4) while eliminating others (C3/C4), adding ambiguity to the insertion loss measurement for the device under test (DUT).

  Figure 2.  Optical test configurations can contain hidden sources of error, such as this insertion loss test setup that does not account for the insertion loss of the new connector pairs it introduces during a measurement. Source: Agilent Technologies.

  
  
• Make estimates on the accuracy of your test equipment and perform error-tolerance calculations to define the margins and guard bands you need around test specifications. Wide guard bands lower production yield, so a good case can be made for the acquisition of high-quality test equipment to keep the margins small.
• Control the testing environment carefully. Unfiltered air contains airborne contaminants that are likely to intrude on test accuracy. Incandescent lights produce significant amounts of infrared radiation, which shows up as noise in optical tests. Vibration and temperature variations can alter test results, as well.
• Use mechanical cleaning. Simply washing an optical surface with alcohol and blowing it dry has been shown to be ineffective in removing surface contamination. In fact, the process tends to "bake in" contamination. Use of a soft, lint-free tissue provides better results.
• Continually monitor your signal sources in order to track out thermal and other environmental drift effects.
• Use free-space optical paths rather than fiber where possible. Handling fibers creates internal stresses that alter the optical path, creating test variations.
• Pay attention to connections. Optical connectors degrade with time and use, so make frequent reference measurements against your calibration standards. In addition, keep test configurations connected up to devices being tested during the manufacturing process to avoid introducing errors by breaking and making the connection at each step.

Minneman suggests that optical test engineers press for the creation of a forum by which test engineers from a range of vendor and customer companies can meet regularly to share test approaches, share innovations, and develop common specifications. In addition, members could use the forum to discuss paths for specification improvement. Such forums have increased the effectiveness and productivity of other industries, Minneman notes, and can do the same for photonics. Minneman also recommends political action in support of funding for government standards laboratories such as NIST to develop improved test methods for optical systems.

Minneman also suggests that executives at optical component companies apply the same strategic emphasis on productivity and process improvement that has kept other industries competitive in the face of low-cost overseas labor. Test engineers can help promote adoption of this approach by demonstrating how process and yield improvements can raise profitability. They may also be called upon to determine where additional testing is needed to monitor process trends.

The photonics industry does not have the body of test experience that the electronics industry enjoys, Williams points out, yet it has had to respond to rapid growth. "The industry is constantly pushing the envelope on optics, and test procedures haven't had a chance to catch up." Still, improvements are possible if test engineers rise to the challenge of raising their industry's standards.