In situ swept-wavelength characterization enables higher throughput

Kathryn D. Li Dessau

Aaron Van Pelt

Steve Cason

Greg Smolka

With the growing complexity of WDM and DWDM network systems-more channels, narrower channel spacing, and wider wavelength range-it has become increasingly important to fully characterize the wavelength-dependent optical properties of network components. Properties such as wavelength flatness and polarization-dependent loss can all adversely affect overall network system performance. Devices must be characterized at the channel of interest and across the entire wavelength band. Moreover, because many devices are trimmed during the manufacturing process, a method that provides fast, real-time feedback can significantly improve manufacturing throughput by identifying real-time in situ capabilities.

One option is the swept-wavelength method for high-resolution wavelength characterization of passive fiberoptic components. Compared to the traditional step-and-measure method, it significantly reduces measurement times by at least a factor of five-from hours to just minutes.

How to build the system

Building a swept-wavelength system is simple. Just like the step-and-measure system it consists of a tunable laser source, device under test, optical receivers, and a data-acquisition system. It differs from the traditional step-and-measure system only in that the specific requirements of the source and receiver are different. Instead of "stepping" the source through the wavelength and dwelling for a time at each step for the data to be taken, the wavelength is swept continuously at a constant rate while the output is recorded. This method provides a very high-resolution spectral picture-fast enough to observe changes as a component is being adjusted-and it delivers the resolution of a laser with the speed of an optical spectrum analyzer.

The first and most important component in a swept-wavelength setup is the tunable laser source. It also represents the main difference in equipment between the step-and-measure and swept-wavelength methods. A few parameters are important in choosing a tunable laser source. The source must have a very linear, repeatable, and mode-hop-free scan to have good relative wavelength accuracy.

There must also be a stable low-jitter trigger. In a sweep, an electrical trigger signal alerts the data-acquisition system, whether it is a computer or an oscilloscope, to begin taking data. This step is key to determining the relative and/or absolute wavelength accuracy. Wavelength repeatability from scan to scan is also important if anyone is watching in real time. Any inaccuracy will cause the data to shift with respect to wavelength from scan to scan, and determining real-time changes from spurious shifts will be difficult.

Finally, a swept-wavelength source must have good output-power stability, both power flatness within the scan and power repeatability from scan to scan. Power flatness within a scan translates to how constant the output power is during a single wavelength scan and is important when looking at a single scan. Power repeatability from scan to scan is a measure of how stable the output power remains during the time the laser makes many wavelength scans. This is especially important when checking the process in real time because any changes in power from scan to scan can look like changes to the device rather than fluctuations from the laser.

The next component required in the swept-wavelength system is the detector (or detectors if monitoring multiple outputs). Indium gallium arsenide (InGaAs) detectors can be used as long as they have enough bandwidth to capture the sweep of the laser, which is typically at least 200 kHz of bandwidth. Calibrated power meters in general are too slow to use. To measure absolute rather than relative power, the InGaAs detectors must be wavelength-calibrated to a traceable power meter.

For multiple-output systems, the best process is for people to build their own detector banks with fiber-coupled InGaAs detectors followed by low-noise logarithmic amplifiers. The output from the amplifier or amplifiers can then go directly to an oscilloscope or data-acquisition system. Because the detectors may have some sensitivity variations over the wavelength scan, output should be normalized using a reference. This step involves breaking away part of the input power to another detector and normalizing.

If it is important to record and store the data, the data-acquisition system should consist of a data-acquisition board followed by a PC. An A-D converter will be needed, and it is important to use a log amp followed by an A-D conversion instead of leading with the A-D conversion first. For a signal of -10 dBm, for example, the dynamic range resolution on the low end is only -0.15 when a linear signal is converted to a logarithmic signal using an 8-bit A-D resolution. This is the resolution when a linear A-D conversion is followed by a digital log conversion.

Data acquisition and analysis

For relative measurements, the real-time, high-resolution data can be viewed simply by using oscilloscope equipment (see Fig. 1). In this example, the laser is programmed to sweep over the wavelength range of interest while providing an electrical wavelength trigger. The optical output is sent to the device under test and then to a detector. The output from the detector is displayed on an oscilloscope and triggered using the laser wavelength trigger output. The stability of the wavelength trigger and the repeatability of the laser scan ensure that motion seen on the oscilloscope display is due to the spectral change of the DUT.

If it is important to store or normalize the data, a computer will be needed to gather data while the laser sweeps across the wavelength range of interest. In this case, the electrical trigger will signal the computer to begin data collection. In the swept-wavelength method, the laser sweeps the wavelength with respect to time so the x-axis indicates time on the oscilloscope. The following equation can be used to convert this feedback to wavelength (see Fig. 2): wavelength = wavelengthstart + sweep speed • time duration from beginning of sweep.

The best way to eliminate intensity inaccuracies is to monitor the power simultaneously during the measurement by using a coupler or splitter to measure a portion of the output power directly and then normalizing the data to that real-time reference. However, if this isn`t possible, the laser output power can be calibrated to a power meter and the output of the detectors measured. A lookup table should be prepared to allow comparison between the output from the power meters and the detectors.

One important note is that the detector sensitivity may not necessarily be flat over the entire wavelength scan of the laser. Thus it is important to calibrate the output of both the laser and the detectors. The best method is to normalize the data with respect to a reference. Additionally, the response of any couplers or splitters, and so forth, must be added to the overall calibration.

Saving time

As an example of the time savings resulting from the swept-wavelength device, consider measuring the throughput of an interleaver product using a traditional step-and-measure system consisting of a tunable laser, device under test, and power meters. For a 100-GHz interleaver with an input spacing of 100 GHz (0.8 nm) and an output spacing of 200 GHz (1.6 nm), a typical measurement from 1525 to 1565 nm in 0.01-nm steps would require 4000 wavelength points.

Using the traditional step-and-measure setup, the data would require more than two hours to gather. With the swept-wavelength system discussed in this article, however, this same measurement requires less than a second (see Fig. 3).

FIGURE 3. Using a swept-wavelength system as an in-process test for alignment, a manufacturer can test interleavers in less than 1 s with data update rates reaching 2 Hz.