Automated Measurement Methods For Fiber-Optic Component Testing

Accurately monitoring optical power levels is essential during design verification or qualification testing of fiber-optic components such as connectors, patch cables and couplers. Most test professionals use one of two methods for multichannel monitoring of optical power levels during these tests, which may last from a few days up to several months. Which method you select will depend on whether you’re doing multimode or single-mode component testing.

Optical power measurements allow you to calculate insertion loss, return loss and change in transmittance as the test sample is subjected to various mechanical and environmental stresses. In the first method, a discrete photo diode is used for each channel; in the second, an optical power meter is coupled with two multiple-port fiber-optic switches.

Using Discrete Photo Diodes

Hardware Considerations

Using a discrete detector for each channel is particularly desirable when testing multimode components due to the potential problems associated with using multimode switches. Any change in modal content of the light passing through the switch can cause variations in loss, which leads directly to measurement errors. Test-system configuration is simplified by the fact that return-loss measurement typically is not a requirement for multimode.

You could provide a separate optical power meter for each channel, but this leads to unnecessary expense due to the redundancy of using multiple displays, chassis, computer interface circuits, front-panel controls and other elements. A more economical approach would be to design a system with multiple detectors and amplifiers multiplexed onto a single or several analog-to-digital (A/D) converter(s) with a single computer interface. One such system that has been developed consists of two 18-channel modules, each containing two PCBs with nine discrete 2-mm active area InGaAs photo diodes (Figure 1), for a total of 36 channels. Each detector is coupled to a programmable gain transimpedance amplifier (PGTA) designed for low offset voltage and input bias current to minimize photocurrent-to-voltage conversion errors.

Although plug-in I/O boards are available with A/D converters, you can obtain a more accurate, lower-noise design by performing this operation as physically close to the amplifiers as possible. In this way, sensitive analog signals are not subjected to voltage drops through the interconnecting cables and externally induced noise from other equipment. All communications between the modules and the PC are accomplished through digital TTL signal lines using a common plug-in digital I/O board.

To meet the resolution requirement of 0.01 dB over nearly six decades of optical power, a 12-bit A/D converter is appropriate. Key parameters include low unipolar offset, low calibration error and a high degree of linearity. Fast conversion time is not particularly important, since even a modest conversion time of 25 m s is more than two orders of magnitude faster than the settling time of the PGTA, which ranges from 10 ms to 400 ms.

Software Considerations

Since none of the optical power modules has a built-in user interface, the software program is the only means of communicating with the hardware. A good choice for this function is a graphical programming language optimized for data acquisition and instrument control, because of its versatility and intuitive graphical user interface. Programming is simplified by the fact that no handshaking is required between the modules and the controlling computer.

You can use a number of software techniques to enhance system performance by maximizing speed and improving accuracy. For example, by storing the gain range for each channel during each scan, the hardware is returned to the previous gain setting during the subsequent scan to determine if the reading is still in range. If it is, then a valid reading can be obtained and no further gain ranging is necessary. This method increases the scan rate considerably because changes in gain range occur very infrequently during long-term component testing.

Another method for increasing scan rate is to allow the operator to select which channels are to be scanned, avoiding needless sampling of dark channels. Still another method is to allow the gain to be switched in both the up or down direction, based on whether the reading is under or over range. This avoids switching through the highest gain ranges, which have settling times of 400 ms, as opposed to the lowest range, which is only 10 ms. By using these techniques, you can obtain typical scan times of 12 s for 36 channels.

Another technique that improves the accuracy and repeatability of the measurement involves acquiring 10 readings for each sample and then computing the average to determine the final reading. This has very little impact on the sample time because the time required to take the readings is relatively small compared to the settling time required for gain switching.

Test Results

When you are measuring optical power levels over long periods of time, two parameters in particular have the potential to influence your measurement result: linearity and stability. Nonlinearities in the measurement system are a direct source of error when you are making relative measurements such as insertion loss or change in transmittance. Stability is of equal importance because a given test can require weeks or months to complete, and any hardware-related signal drift over time is indistinguishable from actual changes in the test sample.

By using the design rules outlined in this article, you can achieve a maximum deviation from linearity of ± 0.04 dB. This compares favorably with the TIA specification of 3%, which equates to ± 0.066 dB (FOTP -20 Measurement of Change in Optical Transmittance). A long-term stability of ± 0.02 dB at 25 ± 3° C minimizes the effects related to system-induced measurement drift.

Using Fiber-Optic Switches

When you are performing tests on single-mode components, the discrete photo diode approach becomes impractical due to the need to measure reflected power for each component. High- performance multiple-port single-mode switches are well-suited for this application because they avoid the pitfalls associated with switching multimode signals.

Although component insertion loss is simply defined as the ratio of output power to input power, and return loss is defined as the ratio of reflected power to incident power, when you are measuring these parameters with a multichannel switch-based system, a number of factors can influence your measurement. By considering these factors and incorporating the proper test procedure, you can obtain accurate insertion loss and return loss values.

Test Procedure

One system configuration suitable for measuring insertion loss and return loss of single-mode components is illustrated in Figure 2. The system incorporates two multiport single-mode switches, a dual-wavelength laser source and an accurate optical power meter. Switch 1 contains an internal bidirectional coupler, a 1 x 2 switch for wavelength selection, and a reference reflector. You can obtain throughput power measurements for determining insertion loss by setting Switch 1 and Switch 2 to the same channel. You can get reflected power measurements for determining return loss by setting Switch 1 to the DUT channel and Switch 2 to the reflectance port channel so that the reflected signal is routed back to the power meter photodetector. Instruments are controlled via the GPIB by using a Windows-based data acquisition package.

When attempting to characterize high return loss (>45 dB) components, you have to fusion-splice the test samples into the system because the reflectance of additional connectors in the signal path would tend to mask out the reflectance of the test sample. Although some of this effect could be canceled out by placing tight mandrel wraps in the cable downstream from the interface connector, this technique does not lend itself well to an automated test.

Since fusion-splice loss is not subtracted out until the end of a test, when cutback is performed, you should establish an estimated splice-loss (ESL) limit to ensure that poor fusion splices are not accepted. Since most fusion splicers provide a calculated splice loss based on the alignment of the two fiber cores, it is a good idea to perform a study which compares calculated to actual splice loss to determine the accuracy of the estimate. This will avoid any surprises at cutback time.

Cable Management

Cable management is an integral part of the test procedure. You can keep any polarization-dependent losses in the signal path to a minimum by maintaining the state of polarization through the cables as much as possible, minimizing stress and physical disturbances once testing is underway. This is achieved by laying the switch pigtails out parallel to one another (not overlapped) and keeping the test system and cables as stationary as possible throughout the test sequence.

For insertion-loss measurements, all readings can be normalized to a source monitor channel, which is a continuous cable spliced between each switch with no test sample in the signal path. This technique cancels out any drift in the output of the laser source. BellCore recommends taking the average of four such monitor channels to determine the small amount of nonrepeatability associated with the optical switch (BellCore TR-NWT-000326 Generic Requirements for Optical Fiber Connectors).

During return-loss measurements, the power meter measures the combined reflected light, which includes contributions from the test sample and the measurement system itself. Since the meter has no way of differentiating between these two components, you need to measure the system background reflection before the test samples are inserted by splicing the switch pigtails together and measuring the reflected power. Once the test samples are inserted, subtract the background reflection to find the true return loss. You can achieve the highest degree of accuracy by maintaining a 10-dB margin between the background reflection and the reflectance of the test sample. Most of the background reflection is due to the internal bidirectional coupler and the Rayleigh backscatter from the fiber pigtails. By keeping the pigtails under 10 m, background reflections on the order of -65 dB are obtainable.

All return-loss readings can be normalized to the value obtained from a reference reflector built into Switch 1. The operation is similar to using the source monitor for insertion loss measurements, except that the return-loss reading is normalized to a signal that has traveled through the reflectance path.

To calculate the return loss from the absolute reflected power, you need to know how much loss the reflected signal encounters as it travels through the reflectance path. Splice a known calibrated reflector to each output of Switch 1 and measure the reflected power. The loss can then be determined as the incident optical power minus the calibrated reflector value minus the measured reflected power. With these techniques, you can achieve insertion loss repeatability of 0.05 dB and return loss sensitivity of 55 dB.

About the Author

Daniel Brown is a Product Development Engineer with AMP’s Optical Interconnection Systems Group. He joined the company in 1992 as a Test Engineer. Previously, he designed test equipment at Meson Fiberoptics in Binghamton, NY. Mr. Brown received his bachelor’s degree from Binghamton University. AMP Inc., Branchburg Corporate Center, 61 Chubb Way, Somerville, NJ 08876, (908) 685-2000.

November 1995