Wavelength-division multiplexing links mandate complex spectral measurements
Common optical test instruments need computer assistance to execute the rigorous component tests to ensure wavelength-division multiplexing operation
Because wavelength-division multiplexing (WDM) technology is moving rapidly from research into production, new and accurate test procedures are needed to supplement established fiber-optic measurement practices. In a WDM network link, the installed components and devices must be characterized based on the proper handling of four or more parallel wavelengths.
This characterization demands the integration of existing fiber-optic test instruments because a specialized WDM maintenance test set is not yet commercially available. For example, a broadband light source, such as an edge-emitting light-emitting diode or a tunable laser source, is needed to replace a light source that works at only one wavelength. In addition, to accommodate multiple wavelengths, an optical spectrum analyzer (OSA) must be implemented to detect spectral information that cannot be resolved by a power meter. Furthermore, the number of WDM channels increases the number of measurement ports on some components. Consequently, to handle that requirement and to automate numerous test procedures for labor, time and cost efficiencies, optical switches and a computer are also included.
A typical WDM system consists of several laser transmitters operating at different wavelengths. Today, many fiber-optic networks use distributed-feedback lasers operating at 1550 nanometers. Each laser supplies one WDM channel wavelength. The wavelength difference between adjacent channels is generally approximately one-half to a few nanometers, depending on the number of channels and the available wavelength range of the link.
Coupled to the laser transmitters, a multiplexer combines the multiple optical signals before they enter the fiber-optic link. When the WDM network span between transmitters and receivers exceeds 50 kilometers, optical amplifiers can be inserted to compensate for signal transmission losses in the fiber-optic cable. At the receiving end of the link, a demultiplexer separates the different optical carriers before the receivers convert them into electrical signals.
Because of the carrier complexity of WDM network design, the following measurements (and possibly others, such as the modulation bandwidth of the transmitters or receivers, or the receiver sensitivity) should be made at the component or subsystem level:
Output spectrum of the distributed-feedback laser and its dependence on backreflections
Insertion loss and return loss of the multiplexer
Insertion loss, crosstalk and return loss of the demultiplexer
Polarization dependence versus wavelength of the multiplexer, demultiplexer and other connected passive components.
The distributed-feedback semiconductor lasers provide a main signal (the desired optical carrier), sidemode signals and spontaneous emissions. The spontaneous emissions are much smaller in output level than the sidemodes and can be ignored in this test analysis. However, sidemodes (that is, the unwanted wavelengths at which the distributed-feedback laser slightly oscillates) can occur at wavelengths used by the other WDM channels and can still cause crosstalk problems.
Another parameter requiring test is the sensitivity of the distributed-feedback lasers to backreflections. If one or more components along the transmission line reflect some light back into the chip, then the output power or the output spectrum can become unstable. To avoid this problem, many distributed-feedback lasers implement an internal or external optical isolator.
Comprehensive distributed-feedback laser tests must accurately determine the center wavelength of the laser, the linewidth, the output power, the sidemode suppression ratio (the difference between the main signal and the highest sidemode, expressed in decibels), and the ability to sustain backreflections. The core instrument used for these measurements is typically an OSA, which is often used with a current source (to drive the distributed-feedback laser) and an optical attenuator (used as a variable backreflector).
Optical spectrum analyzers presently contain built-in application programs to test distributed-feedback lasers quickly and accurately. The transmitter is connected to the input of the OSA and then activated. The analyzer measures the spectrum of the laser; its internal program calculates the center wavelength, linewidth, sidemode suppression ratio, output power and other parameters of interest. Next, the measured values are displayed on the analyzer`s front-panel screen.
To test the laser`s sensitivity to backreflections, a 3-dB coupler and an optical attenuator set for backreflection mode are added to the test setup. The attenuator and the OSA are connected to the two outputs of the coupler, and the distributed-feedback laser and a power meter are connected to the coupler`s input.
The attenuator reflects a variable amount of light back to the coupler. The coupler splits half of this light to the laser and the other half to the power meter, which tracks the amount of reflected light. The OSA monitors whether or not the transmitted spectrum changes with the attenuator setting and, therefore, with the backreflection. To ensure high accuracy of absolute power levels, the OSA must first be calibrated against a traceable power meter, and the insertion losses of the coupler must be characterized and incorporated into the measurements.
The multiplexer combines several input signals to one common output. To couple most of the incident laser power into the WDM link, the insertion loss of the multiplexer for input signal n (which is connected to the transmitter with the wavelength of channel n) should be low at that wavelength. For all other wavelengths, the insertion loss at input n can be either high or low, because all the other wavelengths are not connected to this input. Therefore, the insertion loss is tested only at the specified wavelength for each channel. Consequently, the test result is a one-dimensional vector: Element k represents the insertion loss of channel k when the matching wavelength is applied to input k.
A typical test setup to characterize the multiplexer consists of a broadband source (such as an edge-emitting LED), a 1-by-n optical switch, an OSA and a computer to control the entire test procedure. The switch first connects the source directly to the OSA for normalization. Then it applies the source to all the inputs of the multiplexer. After each step of the switch, the OSA sweeps a trace to measure the signal loss versus wavelength. Next, it sets a marker on the trace at the wavelength of the active input channel. The amplitude readout of this marker minus the amplitude measured at that wavelength during normalization represents the loss of the multiplexer for this channel.
The optical switch can also help determine the return loss of the multiplexer. In this case, a return-loss meter is connected to the input of the switch. Again, the switch outputs are connected to the inputs of the multiplexer. However, an additional output must be connected to a known reference connector; for example, a gold-plated connector reflecting more than 98% of the incident light. Furthermore, the OSA must be replaced by a good termination--one with at least 10 dB more return loss than expected from the device under test. The termination eliminates the measurement error that a reflection behind the multiplexer can produce.
Demultiplexer losses and crosstalk
The demultiplexer separates the incoming optical carriers to provide only one channel to each receiver. The insertion loss for channel n at wavelength n should be low, but it still must be high enough in value to rebuff all the other carrier wavelengths to minimize crosstalk. Therefore, the demultiplexer characterization consists of n insertion-loss measurements plus n times (n-1) crosstalk measurements.
A special test setup for automated demultiplexer testing consists of an edge-emitting LED, a switch and an OSA. The LED typically provides more than -37 decibels relative to milliwatts/nm of output power. This level allows the efficient characterization of components that insert up to a 40-dB loss. However, in some test cases, the edge-emitting LED must be replaced with a tunable laser source if the channel spacing is less than 1 nm or if the insertion loss plus crosstalk suppression exceeds 40 dB. The advantage of using the tunable laser source in the channel spacing test is that its tuning resolution is more than 10 times finer than the resolution bandwidth of the OSA. The laser source can also provide more than a 2-dBm output power; therefore, even in the insertion-loss test, devices such as a two-stage isolator can be tested efficiently for isolation.
An edge-emitting LED source can provide a broadband spectrum. As usual, the test setup must first be characterized without the device under test. Therefore, the light source must be subsequently connected with a patch cord to all switch inputs.
With the use of an edge-emitting LED, the OSA can make a single sweep for each switch state. Then it moves a marker along the trace to all carrier wavelengths to measure the power level at these wavelengths. Including the switch steps, the result is an n¥n matrix where each element (k, l) contains the source power measured at switch input k and at wavelength l. If a tunable laser is used, then it is necessary to tune the tunable laser source to each carrier`s wavelength and perform another sweep before a marker can be used to obtain an accurate readout.
After the source and switch characterizations, the device under test--the demultiplexer--is inserted between the patch cord and the switch. For high accuracy, it is important to keep the source-to-patch cord connection unchanged. Again, the switch selects one output channel at a time, and the OSA sweeps and places markers using the same procedure as before. This procedure leads to a second matrix containing the demultiplexer output power for output k at wavelength l in element (k, l).
To calculate losses, the first or source power table must be subtracted from the second table. The results contain the insertion loss in the diagonal (k = l) elements and the suppression of wavelength k to output l (k is different from l) in all the other elements. The crosstalk of wavelength k to channel l can be calculated as the difference between elements (k, l) and (l, l).
If the return loss of the demultiplexer is a critical parameter, then it must be accurately measured. Any open output of the demultiplexer can have a 4% backreflection. Therefore, all outputs must be either properly terminated or connected to devices with defined return-loss properties (such as the receivers used in the link).
Although a demultiplexer has only one input connector (to which a return-loss meter is usually connected), it is important to use a tunable light source for the return-loss measurement. Because it is wavelength-selective, the demultiplexer could exhibit an acceptable return loss at one wavelength and an unacceptable value at another.
Lasers provide highly polarized light. Because of very small mechanical or temperature changes in a singlemode fiber-optic cable, the state of lightwave polarization is not constant. If the loss of an optical component depends on its state of polarization, the power of each WDM channel will fluctuate over time.
Fortunately, there is a convenient way to characterize the polarization-dependent loss over wavelength for passive components such as multiplexers and demultiplexers. The OSA needed for this test must have a built-in monochromator output, a photodetector input and a white light source.
The monochromator first filters the white light source so that only a fraction of the spectrum is emitted to the linear polarizer. This polarizer creates a high degree of polarization (that is, almost all the light is polarized and only a small fraction is still randomly oriented). However, the polarization controller randomly changes the state of polarization but not its degree or the power level of the light.
To test the polarization-dependent loss as a function of wavelength, the OSA sweeps the monochromator several times. It also measures the power at the output of the device under test that is connected to the photodetector input of the OSA. Then, a sophisticated signal processing function is executed that uses three different traces on the OSA screen: Trace A displays the received power; Trace B captures the maximum power of subsequent sweeps and Trace C captures the minimum power.
Because the polarization controller and the sweeping monochromator apply a variety of states of polarization and wavelengths over a set time, one of the many states of polarization applied at each wavelength will eventually occur at the minimal polarization-dependent loss value of the device under test; this maximum power loss is captured as Trace B. Similarly, another state of polarization will eventually occur at the device`s maximum polarization-dependent loss value; this minimum power loss is captured as Trace C. The loss value is calculated as the difference between Traces B and C. u
Joachim Vobis is product marketing manager at the Hewlett-Packard Co. Lightwave Operation in Santa Rosa, CA.