The rapid growth of wavelength-division-multiplexing (WDM) technology is putting pressure on component manufacturers to design and fabricate improved wavelength-specific components, as well as to develop rigorous testing protocols to monitor production and final quality control. To be economically viable in this cost-sensitive market, any such tests must be automated and performed at high speed.
Testing methods for passive components fall into two categories. The first are those that use a tunable narrowband source to successively scan the entire transmission band of interest. The second method uses a broadband source in conjunction with wavelength-resolved detection, usually by an optical spectrum analyzer (OSA).
Although a tunable source is preferable in some applications, there are two significant advantages to using a broadband source. First, the source itself can be a relatively low-cost device; in contrast, for 100-GHz channel spacing, the only viable tunable sources at this time are more-expensive external-cavity laser diodes or, more recently, a filter-tuned erbium-doped fiber amplifier (EDFA). The second advantage is measurement speed: a broadband source tests all wavelengths simultaneously, whereas a tunable laser must slowly scan the relevant spectral range. In addition, a broadband source eliminates any potential for interference artifacts stemming from the long coherence length of external-cavity laser diodes.
The ideal broadband source for this purpose should have a flat (or smooth and predictable) power output over a wide spectral bandwidth and be able to launch high optical power (at least 100 µW) into a single-mode fiber. This latter characteristic is important because higher power translates into faster data accumulation at a given signal-to-noise level. Finally, to be competitive with tunable source methods, the broadband source should be an inexpensive, commodity item because its deployment will generally require the expense of some type of OSA instrument.
In the 1550-nm spectral region, there are a number of potential sources for this application. For example, a tungsten lamp offers spectrally smooth, broadband output and can easily generate many watts of power. However, its spatially diffuse output cannot be effectively coupled into a fiber. On the other hand, an edge-emitting light-emitting diode (LED) can be readily coupled into a fiber and has a spectrally smooth output. But the LED offers very low spectral power density (power per unit of spectral bandwidth) and cannot support high-speed testing. A better alternative is the EDFA operated in a free-running, amplified spontaneous emission (ASE) mode. However, this option, which combines smooth, broad output and high power in the fiber, is not a low-cost source.
Another potential source is the superluminescent-diode (SLD), which combines the output power of a simple laser diode with the broadband emission characteristic of a conventional LED. Moreover, this output can be efficiently coupled into a fiber. The SLD is designed to operate in a stimulated emission mode, analogous to a laser. There are, however, no cavity mirrors (reflective facets) to create a laser cavity. The SLD output is intense because of stimulated emission, and the power/current slope is similar to that of a laser. The lack of multiple cavity reflections results in broadband output rather than a few sharp cavity modes as with a true laser device.
Several factors have prevented the SLD from delivering its full potential for WDM component testing. Most SLDs are simply laser diodes with multilayer antireflection coatings applied to the front and back facets. Fully suppressing the cavity mode structure and achieving spectrally smooth output requires an antireflection (AR) coating with ≤10-5 reflectivity (≤-50 dB optical feedback). Multilayer AR coatings are difficult to create on a laser diode with this level of performance and do not lend themselves to high-volume, high-yield production.
Thus, this design approach yields either a higher-cost device with truly smooth spectral output or a lower-cost device (with an AR coating in the 10-3-10-4 range) that exhibits periodic ripples that will drift in the wavelength domain with even slight changes in device temperature (see Fig. 1). This rippling becomes more pronounced at higher operating current (higher output power). Neither is a perfect option for this application.
An alternative approach is an SLD designed to operate as such from the outset rather than as an AR-coated laser diode. The design of the device includes a gain region that does not extend completely to the facets-there is a "dead" region between the active, index-guided junction and the facets. Because of the natural divergence of light as it leaves the index-guided gain stripe, it overfills the output facet, reducing the effect of reflective feedback from this facet. As this reduction also lowers the power that will be coupled out of the facet, the coupling distance must be carefully optimized.
The device can be operated at high current without the risk of lasing; 0.5 mW of optical power can be launched into the fiber. Such a device has a smooth spectral output, with a spectral bandwidth as high as 50 nm at the 50% (-3 dB) point. For WDM testing, this SLD can be fabricated with a center wavelength anywhere in the 1550-nm wavelength region.
Broadband light sources based on this SLD design typically combine both variable power and modulated (chopped) output. One of the most common applications of a broadband light source is to perform high-speed, wavelength-dependent loss measurements of fiber Bragg gratings, as well as subsystems such as an assembled wavelength-division multiplexer (see Fig. 2).
The broadband light source is connected to the OSA via a length of low-loss "reference" fiber. All insertion losses are then measured with respect to this input source measurement. The device under test is then inserted between the source and the OSA. The observed spectral density (in decibels) is subtracted from the input source measurement to obtain a plot of insertion loss versus wavelength. This type of normalized measurement compensates for the fact that the spectral density of the broadband light source is wavelength dependent. While there is still wavelength dependence in the measurement noise floor, it is not usually a problem with a true SLD source (see Fig. 3).
The wavelength accuracy of any measurement using an OSA is only as good as the absolute wavelength calibration of the OSA itself. An SLD-based broadband light source can be used to periodically recalibrate an OSA. To do so, the output of the source is passed through a glass reference cell containing isotopically pure acetylene or hydrogen cyanide gas before being coupled into the OSA. Both acetylene and hydrogen cyanide have a simple yet dense spectrum of absorption lines, collectively covering the 1510- to 1565-nm spectral regions. Furthermore, researchers at the National Institute of Standards and Technology (Gaithersburg, MD) and elsewhere have calibrated all these lines with high precision.
Absolute calibration of the OSA, therefore, requires only correlating the observed spectral lines with their known positions. Care must be taken to allow for the refractive index of air; the position of any molecular spectral line is only constant in the vacuum wavenumber domain (cm-1), but the OSA is a wavelength-sensitive device with measurements affected by the refractive index of air. Tabulated vacuum wavenumber values are converted to observed air wavelength measurements where air wavelength (nm) = 107/(vacuum wavenumber × air refractive index). Tabulations of the refractive index of air as a function of temperature, pressure, and humidity are readily available from a number of sources.
This SLD is also a good source for measuring the polarization mode dispersion (PMD) of fibers using white-light interferometry. In addition to quality-control measurement on new fiber, PMD is a particularly important measurement in determining whether older fiber infrastructure can support faster transmission rates. Finally, the SLD may be useful for related tests to measure the polarization-dependent loss of system components.
Yetzen Liu is president of Fermionics Lasertech, 4555 Runway St., Simi Valley, CA 93603. He can be reached at 805-582-0155 or email@example.com.
FIGURE 1. As sources for testing multiplexing components, an LED offers a low-power spectral output, a laser diode has a narrow band of high-power spikes, and an antireflection-coated superluminescent diode (SLD) produces a wider but varying output. A "true" SLD is a spectrally smooth source.
FIGURE 2. A reference fiber links the broadband light source (SLD; see inset) in the test setup to an optical spectrum analyzer for measuring a baseline instrument reading. Once that measurement is known, a device to be tested is put in place of the reference fiber.
FIGURE 3. In a spectrum of a fiber Bragg grating demultiplexing filter, obtained from an SLD and an optical spectrum analyzer, a reference fiber measurement is subtracted from the observed data to yield a plot normalized for source power.