Therefore, measurement and characterization procedures for DWDM optical devices and subsystems are necessary, as well as analyses of the optical network's operating parameters, to ensure the quality of service expected by the user. This article will analyze the main parameters and measurement procedures for DWDM optical devices, leaving the analysis of operating parameters and the main measurement equipment in DWDM optical networks for a subsequent article.
Test measurements on DWDM devices
When analyzing the quality of optical components, whether for R&D, production lines, or maintenance purposes, comprehensive and rigorous measurement tests are critical. Therefore, design and product engineers must have in-depth knowledge of the measurement processes for key optical parameters of the components, as well as the interpretation of the measurement results. The main optical parameters that can be measured in a DWDM device will be briefly discussed below.
Insertion losses
As with any electronic device, insertion loss refers to the difference between the power (in this case, optical) entering the device and the power leaving it. Ideally, this value should be as small as possible. It is defined for a specific wavelength of the optical signal. Therefore, in the case of DWDM devices, it is necessary to specify this value for all wavelengths of the operating channels. It's important to note that insertion loss is defined for passive DWDM devices such as filters, couplers, or multiplexers, while for active devices, such as optical amplifiers, the gain—its inverse—is defined.
The manufacturer typically provides a graph of the device's losses as a function of the optical signal's wavelength for the entire bandwidth where the component is intended to be used. This graph specifies the wavelengths of the DWDM channels according to the ITU (International Telecommunication Union) grid, along with their allowable tolerance due to variations in the transmitter's wavelength. The best way to define insertion loss is then to refer to the highest loss (worst-case scenario) within the relevant ITU channel. For example, Figure 1 shows these parameters for a DWDNM channel filter. Finally, when specifying insertion loss, it is important to know whether the device was measured with or without connectors.
Curly
As shown in Figure 1, the passband of a DWDM device is not perfectly flat. The insertion loss value reflects the device's losses at a given wavelength, but it provides no information about the variation of these losses across the passband or the ITU channel. Therefore, the ripple parameter has been defined to characterize this variation with wavelength. Ripple is defined as the difference between the maximum and minimum insertion loss over the passband, as shown in Figure 1. Knowing the ripple value allows us to estimate potential variations in the transmitted optical power and verify whether the device meets its design specifications. In most devices, high ripple values are unacceptable.
Another directly related parameter is the maximum ripple slope. In this case, it refers to the variation in channel power resulting from small changes in the transmitter's wavelength. It is specified in units of dB/nm, and its value can render a device unsuitable even if it has low insertion and ripple losses.
Bandwidth
The bandwidth of an optical device, as with any electronic component, is defined as the spectral width for which the device's response exceeds a certain value (typically -3 dBm of its maximum or FWHM value, see Figure 1). To define the bandwidth, it is essential to specify this threshold, with typical values being 0.5, 1, 3, or 20 dB below the maximum. The bandwidth of a device serves two purposes. First, it identifies the spectral range where the device exhibits acceptable loss values (around 0.5 or 1 dB), that is, what is called the passband of a multiplexer, for example. Second, it defines the selectivity or skirt roll-off of a filter, for example. In this case, the bandwidth value at -20 dB is used. Thus, an ideal rectangular filter will have identical bandwidth values at -0.5, -3, and -20 dB. Manufacturers are beginning to use a new parameter known as the figure of merit (FOM) to describe the shape of a filter's response. This parameter is defined as:
FOM = bandwidth at -3 dB
Bandwidth at -20 dB
Ideally, the FOM will be equal to 1, and in general, higher-order filters will have a more abrupt transition band and an FOM value closer to unity.
center wavelength
The center wavelength is commonly used to describe a multiplexer channel and is defined as the midpoint between the upper and lower cutoff wavelengths of the passband (those at which losses increase by 3 dB). It does not necessarily coincide with the peak wavelength (the wavelength of minimum losses). However, for a device with a perfectly symmetrical response, both wavelengths do coincide. The center wavelength is usually calculated as the midpoint between the wavelengths at which losses of 3 dB are obtained.
Directly related to this parameter is the nominal wavelength, which corresponds to the operating wavelength of the device and is typically found on the ITU grid. The optical transmitter will operate close to this wavelength within a certain tolerance; therefore, this is another important parameter to measure in the case of laser transmitters for DWDM systems. The output wavelength of lasers is subject to temperature drift, which must be rigorously controlled (see CONECtrónica No. 40, pp. 14-16).
Optical return losses
When light is injected into an optical component, most of the optical power is transmitted, but a small fraction is also absorbed or reflected. Reflected optical power is an undesirable effect for several reasons. First, it contributes to increased losses. Second, laser transmitters for DWDM are very sensitive to reflected light, as unwanted reflections can degrade laser stability or the system's signal-to-noise ratio. In some cases, reflections can even damage the device itself. Furthermore, reflected light can be reflected back, combining with the original signal and causing multipath interference that leads to demodulation problems. Finally, reflections within an EDFA, especially at opposite ends of the gain block, contribute to worsening the amplifier's noise level.
In fiber optic systems, reflections are caused by Rayleigh scattering and Fresnel reflections. Rayleigh scattering occurs within the optical fibers due to interactions between light waves and the crystal molecules. The intensity of the scattering depends on the signal's wavelength and the crystal composition. Typical values are around -70 dB/m. Fresnel reflections, on the other hand, occur at discrete points where discontinuities appear, such as inside connectors or adapters. These reflections result from air gaps, misalignments, or changes in the refractive index.
Return loss is a positive dB value, as it refers to the attenuation of the reflected wave relative to the incident wave (incident power divided by reflected power), while reflections are quantified with a negative dB value (reflected power divided by incident power). High return loss values are generally desirable, although there are devices where the opposite is true, since in this case reflection is the desired phenomenon. Some examples of such devices are Bragg gratings and mirrors. The experimental setup for measuring optical return loss is quite simple and is shown in Figure 2. The photodetector directly measures the optical power reflected by the device under test (DUT). However, prior calibration of the system using a DUT with a known value is required.
In DWDM devices, return loss must be measured for all wavelengths of interest. This can be done in two different ways: either using a tunable optical source or using a broadband ASE (amplified spontaneous emission) source. In the case of a multiplexer, measurements must be taken at each input, making the measurement process difficult to automate for devices with a large number of channels.
Directivity
The directivity parameter of a multi-input photonic device is a measure of the isolation between its inputs. This parameter is especially important in multiplexers, where, as previously mentioned, reflections towards the system's transmitters must be minimized. Often, it is sufficient to measure the directivity for only a few channels. The setup is very similar to that shown in Figure 2. The only difference is that the optical source is applied to one of the multiplexer's inputs, and the optical receiver measures the power returning through each of the other inputs. Directivity is therefore the difference in dB between the injected optical power and the measured optical power. For devices with a large number of channels, this measurement is also difficult to automate.
Channel isolation and crosstalk
These measurement parameters are quite typical in the case of DWDM multiplexers and channel filters. Isolation refers to the rejection of the power of a specific channel with respect to another. Figure 3 schematically represents the isolation parameter in DWDM devices. Crosstalk, on the other hand, is defined similarly but refers to the rejection of the power of all other channels with respect to a given channel.
Polarization-dependent effects
As discussed previously (see CONECtrónica No. 48), it is virtually impossible to control the polarization state of a signal traveling through a fiber optic network. This fact, coupled with the fact that many photonic components have characteristics dependent on the polarization state of the input signal, necessitates measuring the polarization dependence of certain device parameters to ensure reliable performance and operation. The main polarization-dependent parameters are insertion loss, bandwidth, and center wavelength. Each of these parameters must be measured at different polarization states using a polarization controller. Although ideally they should be measured for all possible polarization states, in practice, a few random cases are sufficient.
The main measurements typically performed are polarization-dependent losses and polarization-dependent wavelength losses. The former refers to the variation of device losses for different polarization states. When added to insertion losses, they provide the worst-case loss for a component at a given wavelength (usually the nominal operating wavelength). Depending on the device, polarization-dependent losses can be measured over a specific spectral range.
Devices are typically designed to have low polarization-dependent losses. However, there are some exceptions, such as polarizers. Generally, polarization-dependent losses are minimal in the device's passband and higher at the edges and in the stopband. Some manufacturers only perform measurements in the center of the channel and at the edges of the passband. Values of less than 0.1 dB in this region are satisfactory for most applications.
Other measurement parameters related to polarization are polarization-dependent bandwidth and center wavelength. Both parameters refer to the changes that occur in the device's spectral response as a result of changes in the polarization of the input signal and can be measured simultaneously. One measurement technique involves measuring insertion loss and polarization-dependent losses as a function of the signal's wavelength and plotting them on a graph. The graph's envelope reflects the possible loss values for different polarization states, from which the variations in bandwidth and center wavelength can be extracted in the worst-case scenarios. Finally, another measurement alternative involves using an optical spectrum analyzer to perform different sweeps for multiple distinct polarization states.
Measurement of the spectral response of a device
As explained previously, determining many parameters of DWDM photonic devices requires measuring the device's spectral response for each optical channel. Several experimental setups are possible for this measurement, which we will discuss below. The choice of setup depends on factors such as spectral resolution, dynamic range, and the number of DWDM channels to be measured.
The first setup can utilize a broadband ASE source that generates a power distribution across the entire DWDM channel range. This source is connected to the DUT, and the device's response is measured using an optical spectrum analyzer. The spectrum analyzer will have been previously calibrated by measuring the output spectrum of the ASE source. The experimental setup is shown in Figure 4. The main limitation of this technique is that the measurement resolution is determined by the analyzer's resolution bandwidth, which is typically around 0.05 nm.
To improve measurement resolution, which is critical for certain DWDM devices, the ASE source is often replaced with a tunable laser with a reduced linewidth. In this case, the resolution is determined by the laser's tuning step, achieving values as high as 0.001 nm with excellent dynamic range. The measurement involves performing a scan with the spectrum analyzer for each laser tuning step. The main drawback of this technique, however, is that it requires more time to perform the measurement, especially in the case of multichannel devices.
Another alternative involves using a tunable laser and several multichannel optical power meters, as shown in Figure 5, each measuring several DWDM channels of the DUT. In this case, the measurement is independent of the number of channels in the device, resulting in very fast measurements. The measurement resolution is determined by the laser source, while the dynamic range is determined by the source's spontaneous emission noise. Dynamic range values of around 60 dB can be obtained. This technique can also be used to measure optical return loss and polarization-dependent losses by slightly modifying the setup shown in Figure 5.
Finally, to further increase measurement speed, the tunable laser can be operated in continuous sweep mode. Wavelength accuracy is controlled with a reference instrument, and a polarization controller is typically used simultaneously to measure polarization-dependent parameters. In practice, this latter configuration is generally preferred.
Author: Francisco Ramos Pascual. PhD in Telecommunications Engineering. Professor at the Polytechnic University of Valencia.
