Measurement techniques for characterizing DWDM passive components are evolving almost as rapidly as the components themselves. The reduction in channel spacing (200 GHz to 100 GHz to 50 GHz-and even to 25 GHz) means that the passive components that combine and separate the modulated light energy must be characterized by ever-increasing resolution.
The tried and tested way of using a broadband light source with an optical spectrum analyzer (OSA) is no longer suitable for high dynamic range, high resolution, and high-accuracy measurements. This limitation is caused by a combination of the resolution bandwidth of the OSA and the spectral power intensity of the broadband source. For this reason, most DWDM passive component testing is performed with a tunable laser source (TLS) and either a power meter or an OSA.
In the TLS / OSA configuration, the measurement resolution is determined by the TLS motor step size and linewidth and not by the OSA. The measurement sequence consists of repeated OSA scans while stepping the TLS across the wavelengths of interest at the desired resolution. Loss curves with a resolution as high as 0.001 nm as well as an excellent dynamic range can be obtained. The main drawback of this setup is the time it takes to perform a scan. Also, for multichannel components, only one channel can be measured at a time.
In the TLS / power-meter configuration, a low-noise TLS scans across the wavelengths of interest, while at the same time multichannel power meters acquire transmission data for each channel being tested. With this type of setup, the measurement time is independent of the number of device channels and is particularly well suited to testing multichannel components (see Fig. 1).
Measurement resolution is determined by the TLS, and the measurement dynamic range is determined by the local spontaneous emission (surrounding the peak) or total spontaneous emission (or noise) level of the source. Optical return loss and polarization-dependent loss measurements can be integrated into the process, as can a wavelength meter used for dynamic real-time wavelength calibration. Additional power meters can be added to cover an ever-increasing number of device channels (see Fig. 2).
Measuring wavelength uncertainty is obviously the most stringent requirement in new WDM filter testing. For 0.8-nm (100-GHz) channel spacing, most system manufacturers believe that an optical filter should be characterized by an accuracy of ±0.02 nm because typical 3-dB bandwidth is ±0.2 nm. This requirement is especially true at specific channel points such as -1 dB and -3 dB bandwidth, the central wavelength, and the peak transmission wavelength. Performing measurements at these three points provides enough confidence about both extreme and typical losses encountered by a specific carrier as it propagates through an optical channel.
Filter manufacturers that wish to use tunable laser sources to test passive optical components for spectral insertion loss need to be concerned about unit calibration. They need to ask whether the intrinsic calibration of the product is good enough to account for all operative wavelengths, optical power, or drive currents, and all environmental conditions.
Another requirement-wavelength linearity-is also known as the relative wavelength accuracy of the output wavelength through the range. Perfect linearity would require every point that is accessible through the tuning mechanism to be associated with a particular wavelength. However, in practical terms, this is impossible for reasons of stability and repeatability. Tuning mechanisms used to shift the laser central wavelength are rarely perfectly linear, and many points must be calibrated to improve wavelength-setting linearity. Does a recalibration at one point positively correct all other points? The answer is usually no.
Finally, wavelength repeatability can be defined as the ability of a source to reproduce the same wavelength after successive scans. The variation between the actual wavelength settings should be reasonably low, at least down to the relative accuracy. Note that this TLS specification cannot be corrected or calibrated because it is a random behavior. Wavelength repeatability is thus very important in choosing a TLS for testing insertion loss or other loss parameters such as polarization-dependent loss or optical return loss through the spectrum.
Testing with an erbium-doped fiber laser
In selecting the central frequency, it is simpler to use an intracavity tunable interferometric filter than the more-common grating-based external-cavity laser. This laser is more difficult to use for two reasons. First, the design of such a laser renders it more sensitive to parasitical Fabry-Perot cavities. Such sensitivity allows side modes to compete, which results in poor power stability. Second, it is difficult to achieve precise and stable relative mechanical positioning of the external optical feedback assembly. Historically, the goal of this design was to achieve a pure single-mode-operation tunable laser for spectroscopic analysis of gas or solid-state materials.
On the other hand, the erbium-doped fiber laser was designed specifically for passive component testing in which optical features and paths are on the order of a few centimeters. It was designed-through the selection of an appropriate intracavity filter-to produce a linewidth of less than 1 GHz (0.008 nm) at -3 dB. This linewidth results in a temporal coherence length of about 10 cm. Doped fiber-based lasers offer a good balance between small linewidth and stable output power, in comparison to the much greater coherence length of external-cavity lasers.
In-line polarization controllers and polarizers are used to optimize interferometry-based filtering (see Fig. 3). The source output is therefore highly polarized. Side-mode noise, which is difficult to avoid in external-cavity laser designs, is barely present. This filtering cuts out all the optical amplitude spontaneous emission (ASE) that may be present outside the transmitted band. An ASE suppression of more than 65 dB has been obtained.
Most external-cavity laser designs based on a rotating assembly (including mirror and gratings) show longitudinal modehopping because they are single-mode laser designs. The erbium-doped fiber laser design offers stable and continuously tunable high-output power (+4 dBm from 1520 to 1570 nm). Also, by using a simple optical switch and cutting the ring cavity, a nonpolarized high-output-power ASE source (broadband source) can be obtained.
A tunable source can be connected to a WDM common port and each channel insertion-loss profile measured. Using a wideband-sensitive photodetector instead of an expensive OSA gives similar results, provided that the spontaneous emission of the TLS is low enough that the power measured is in fact a function of the device loss profile (see Fig. 4).
By inserting the output of a TLS into an erbium-doped fiber amplifier, a very high-power tunable source that produces more than +14 dBm can be created and still obtain a reasonable optical signal-to-noise ratio. Over 40 dB of spontaneous noise suppression between 1530 and 1570 nm as well as very flat output power along the complete tuning range is possible. High power in passive component testing is required for measuring directionality or for measuring high-loss optical paths passing through a double-stage isolator, for example.
Because the rotating filter material expansion coefficient is small, no temperature control is required to obtain long-term wavelength stability, even when the external temperature varies from 0∞C to 40∞C. The very low influence of ambient temperature allows accurate readings at temperatures other than the original calibration temperature.
Source wavelength accuracy can be enhanced using a multiwavelength meter, which can provide as many points as desired combined with an absolute set point uncertainty of ±5 pm. To maintain a high level of confidence, software integration and automated calibration procedures can be performed at regular intervals, such as once a week. Our tests of long-term wavelength stability have shown drift of less than 0.01 nm over 7 hours.
Additional benefits to using an erbium-doped-fiber tunable laser source for testing include good wavelength repeatability and precise return of output power after multiple scans. These properties, combined with its ability to work with a power meter rather than a more expensive OSA, make a strong case for the value of this approach.
I would like to thank Michael Carlson, EXFO application engineer, for his contribution to this article.
Olivier Plomteux is scientific product manager for EXFO, 465 Godin Avenue, Vanier, Quebec, Canada G1M 3G7. He can be reached at 418-683-0211 or firstname.lastname@example.org.
FIGURE 1. A tunable, erbium-doped fiber-laser-based technology used with a power meter can increase confidence in the qualification testing and manufacture of passive components and filters for DWDM.
FIGURE 2. In a TLS / power meter configuration, the tunable laser source scans across the wavelengths and the power meters gather data for each channel under test. Optical return loss and polarization-dependent loss measurements can be integrated into the process.
FIGURE 3. Erbium-doped fiber-laser tunable source offers stable and continuous high-output power of +4 dBm between 1520 and 1570 nm.
FIGURE 4. The same filter insertion loss measured using an affordable highly linear power meter occurs with both an erbium-doped fiber TLS and a commercially available external-cavity TLS.