To advance in parallel with the evolution of the component and network elements, test-and-measurement equipment must meet new requirements. Advanced erbium-doped fiber amplifiers (EDFAs), for example, are tested not only for gain, noise figure, output power, gain flatness, and polarization dependent gain, but also for new features like transient control.
To test amplifiers for transient behavior, a channel-drop event is simulated by abruptly reducing the input signal power, corresponding to the drop in power when one or several channels in a DWDM link are dropped in a dynamic network configuration. From the output response, the amplifier system is characterized for overshoot and undershoot in power and settling time (the time it takes for the power to recover to the original level and the error in reaching this level).
Emerging amplifier technologies based on semiconductor optical amplifiers (SOAs) and erbium-doped waveguide amplifiers (EDWAs) also require new types of testing. At an early production stage, chip-level tests are required before the device is pigtailed and packaged into a module. Because the chips usually do not have fibers attached, positioning and alignment equipment is needed to connect the device to the test setup. Manual processes take time and reproducibility is limited by the uncertainties of human operators. Automated or semi-automated fiber-to-chip coupling is preferable, depending on the required throughput.
The rectangular cross section of the planar waveguide structures in SOAs and EDWAs can also cause an inherent dependence of gain and noise on the state of polarization (SOP) of the input signal. To minimize this dependence, polarization-resolved measurements are important and can reduce test costs. One way to determine the dependence is by sequential measurements with the input signal at the two principal states of polarization, which can correspond to the TE and TM waves. However, the process of finding and setting these polarization states is time-consuming, especially for devices with fiber inputs when the relationship between the polarizations at the two ends of the fiber is unknown and variable.
Simplifying the instrumentation used to determine noise figure helps vendors reduce cost-of-test. The noise figure of optical amplifiers is determined by measuring the amplified-spontaneous-emission (ASE) power and the output-signal power. Two common ways to do this are interpolated source subtraction (ISS) and time-domain extinction (TDE).
In the ISS method, the ASE.power is measured at wavelengths offset from the signal. Curve fitting obtains the ASE power at the signal, using the spectral resolution of an optical spectrum analyzer (OSA). The amplified contribution from the source spontaneous emission (SSE) of the input signal is then determined and subtracted from the ASE measurement to obtain the ASE caused by the amplifier.
The TDE method, also known as the pulse method, is especially adapted to erbium-doped amplifiers and exploits the relatively slow reaction time of the erbium ASE to power changes in an on/off modulated input signal. The power of the ASE is sampled during the signal-off parts of the modulation cycle. Typical modulation rates are 50 to 100 kHz or higher for faster electronic pump regulation.
This TDE method is not applicable to SOAs because of their rapid response time, but is widely accepted for EDFA measurements, in part because of its reduced sensitivity to some instrumentation specifications. Sources with high SSE lead to inaccuracy for the ISS method when the SSE contribution to the measured ASE is dominant. This is a particular problem when the test source needs a booster amplifier in the test setup to achieve the desired power levels, since the booster adds spontaneous emission. By extinguishing both the signal and the SSE, the TDE method is immune to this problem.
Because the ISS method must distinguish low ASE signals from nearby strong signals, the wavelength resolution and dynamic range of the OSA must also be high, while the requirements are again relaxed for the TDE method. Laser sources offer sufficiently low SSE (and often sufficient power to eliminate the need for a booster) and OSAs offer the performance required to obtain highly accurate ISS measurements. Nonetheless, TDE measurements are ideal for comparison with the ISS results, or when other factors play a role, such as the requirement for high input power.
For these cases, TDE capability can be added in a test setup with a minimum of extra instrumentation. Instead of external optical modulators with timing electronics to pulse the source and gate the detection, the sources and OSA can perform these functions directly. A trigger signal from the self-modulating source can be used to control the timing of the built-in OSA gating.
When testing an amplifier for multichannel use, the ability to modulate all of the test sources synchronously is important (see Fig. 1). Built-in measurement routines in the instrumentation enable both types of noise figure measurements with the same instrumentation, offering simplicity and investment savings. Avoiding an external modulator for the sources also enhances the power budget.
The International Electrotechnical Commission (IEC) is preparing two standards for the pulse method of noise-figure measurement in multichannel applications, either using external modulators or this built-in OSA gating. In some cases, external modulators are necessary—for example, to achieve high modulation rates. At high rates, external modulators extend the extinction ratio of the modulation beyond that of electronically gated OSAs. Combining the interpolation function with the TDE routine, however, subverts the need for a high extinction ratio, measuring the ASE at wavelengths slightly offset from the signal. In this way, the low-cost TDE setup can avoid inaccuracy.
A comparison of the noise figures for an EDFA obtained via the TDE and ISS methods shows close agreement (see Fig. 2). Combining the sources with a wavelength-dependent multiplexer, in this case based on thin-film filters, gives a low insertion loss and prevents accumulation of overlapping broadband SSE from the multiple distributed feedback lasers (DFBs). In this way, we confirm the effectiveness of direct TDE and the accuracy of the ISS method, which can also be used for non-erbium amplifiers. The results of the two methods only diverge at high input signal power when the SSE begins to reach a significant absolute power level.
Direct modulation of the test sources, especially synchronously, can also simulate the sudden adding and dropping of input channels to measure the transient response and control of optical amplifiers. Modulation rates of less than 1 kHz are typically of interest as these allow the amplifier time to relax into both the signal-on and signal-off conditions. The rise and fall times of the modulation must be sufficiently fast, in many cases below 100 ns. A sufficiently fast detector can gauge the output response of the amplifier by measuring the time response in sweeps triggered by the modulated source.
Besides the square-wave power modulation used for TDE and transient testing, other modulation forms are useful. Advanced modulation of the laser current or the laser cavity can increase the effective linewidth of the laser source. A shorter coherence-length source reduces measurement fluctuations due to reflections in the device under test or the test setup.
Especially in unfinished SOAs, the reflections at the facets of the chip can result in coherent interference. Detecting such effects is important, and reducing the coherence length can help to minimize their influence on measurements. Similarly, specialized modulation can increase the effective linewidth with respect to stimulated Brillouin scattering (SBS). For high-power signals launched into long fibers, SBS can cause high return loss and high relative-intensity noise (RIN). When testing Raman amplifiers or transmission systems with narrow-linewidth unmodulated sources, SBS can be a significant problem. Appropriate tunable and fixed-wavelength laser sources can suppress the SBS.
Optical amplifiers based on planar waveguide technologies, especially SOAs but also EDWAs, raise the need for testing at specific input states of polarization (SOP) that are aligned with the symmetry axes of the waveguide. As in microwave waveguides, the two polarization states are called transverse electric and transverse magnetic, TE and TM. The gain of an SOA is generally different for these two states. Distinct measurements of them are useful in developing the gain chip, optimizing the final amplifier for low PDG, and characterizing an SOA for polarization-maintaining applications. Indeed, similar needs also exist for passive components based on planar waveguides.
Such measurements are complicated by a fiber connection between the polarization control and the waveguide under test. The relationship between the polarization states at the input and output of this fiber is not known a priori and often not long constant. The settings of the polarization controller must be adjusted by serially searching the settings of the controller for the resulting maximum and minimum device output signal associated with TE and TM. However, this approach is slow and costly.
Instead, the controller position can be found analytically from a few measurements at pre-determined polarization states. This Mueller-matrix analysis can be used either to calculate the gain or power for the TE and TM states directly, as well as the corresponding polarization settings, or to actually make those instrument settings and measure the TE and TM signals directly. Both of these approaches are useful, depending on the test procedure, and both are faster than an iterative search.
Such gain measurements, dependent on input-signal power, agree well with conventional PDG measurements, made by scanning through all polarization states (see Fig. 3). However, conventional measurements of minimum and maximum gain alone could not establish that the minimum at one input power level corresponds to the same polarization state as the maximum at another power level, while the Mueller analysis unambiguously identifies the crossing point of the TE and TM curves.
Dirk Muschert is program manager for optical amplifier test, Michael Kelly is application expert, and Denis Kobasevic is application engineer at Agilent Technologies Deutschland, Optical Communications Measurement Division, Herrenberger Strasse 130, 71034 Böblingen, Germany. Dirk Muschert can be reached at firstname.lastname@example.org.