With the coming of higher and higher channel counts in dense wavelength-division multiplexing (DWDM), interchannel spacing is being greatly reduced—a channel spacing of 25 GHz is currently being demonstrated. In conjunction with this development, system installers are turning more and more to all-optical networks to reduce costs. For such networks to operate properly, extreme accuracy is needed in all the network components.
Most of the multiplexers/demultiplexers (muxes/demuxes) deployed in the field are working far from full capacity. To avoid unpleasant surprises when adding channels—such as discovering that the mux/demux in use is not linear or has other inferior properties—the best approach is to check the real spectral transmittance of a system's passive components when it is out in the field being commissioned.
Two preferred methods can be used to achieve this: having a tunable source and a broadband detector (such as a power meter) and sweeping through the channels of the passive device, or having a broadband source and an optical-spectrum analyzer (OSA) that performs the sweeping. The second configuration is ideal for measuring devices such as filters, isolators, circulators, attenuators, couplers, and other optical components with high loss.
A broadband source such as an amplified-spontaneous-emission (ASE) source used with an OSA can determine bandwidth, center wavelength, ripple, and insertion loss (see Fig. 1). For field use, double-pass grating-based OSAs offer the best signal-to-noise ratio, narrow-resolution bandwidth, excellent dynamic range, and very accurate wavelength accuracy and repeatability.
What to test
One important parameter to be tested is channel count. As channel count increases, the spacing between the channels' central wavelengths is reduced, and the width of each peak is also reduced accordingly. If the central wavelength of a passive device is not exactly the same as that of its corresponding light source, part of the signal may be blocked by the device and some information might be lost.
Peak power and insertion loss should also be tested. Peak power is the main factor that determines how far the signal can go and still be usable. Passive components must have as low a loss as possible on each channel to avoid reducing transmission length.
Information on power flatness and ripple is also needed. Optical amplifiers amplify signals according to their individual power. If all input wavelengths have similar power, all are amplified equally. Having bad power flatness results in odd behaviors in amplifiers.
Signal-to-noise ratio (SNR) is a crucial factor in system performance. The receiving end triggers on SNR level, not on peak power level. Distance and amplifiers degrade SNR. It is important that the passive components do not degrade it too much.
Also to be tested are bandwidth and channel isolation, which help determine the amount of information from one channel that leaks into neighboring channels. As the spacing between channels is reduced, neighboring channels are closer and closer and crosstalk must be controlled tightly.
What an OSA should be
An OSA is a wavelength-agile power meter. For each given wavelength position of the OSA, the instrument must detect all the light at this position but only the light at this position. The filter passband shape should appear as a tall, thin rectangle. The optical rejection ratio specification of the OSA defines the steepness of the filter, while the resolution bandwidth defines the width.
The ASE source is a fairly high-power source. Some passive components have very low insertion loss, while others can have very high insertion loss. An OSA with a good dynamic range allows testing of a broad range of components and component quality.
Filters and passive components have center wavelengths defined by the ITU grid. To avoid crosstalk, each center wavelength has little tolerance. A good OSA should have an accuracy smaller than this tolerance to ensure proper transmission.
When using broadband sources, it is important to understand how the power density affects measurement limitations. Broadband source power specifications include total power, peak power, and power density.
Total power, expressed in dBm or mW, is the total integrated power of the source as would be measured when connecting the output directly to a power meter. Peak power, expressed in dBm or mW, is simply the highest power level across the spectral distribution. For this value to be meaningful, it must be combined with the resolution bandwidth of the OSA. Power density, expressed in dBm/nm, is the integrated power in a .1-nm slice of the spectrum. This is normally measured with an OSA with the resolution bandwidth set to 1 nm. The power density varies as a function of wavelength.
The power density limits dynamic range when performing measurements with a grating-based OSA. For example, consider a source with a relatively flat power density of -10 dBm/nm in an OSA with a resolution of 0.1 nm. With the OSA resolution bandwidth set at 0.1 nm, the reference power would be -20 dBm (-10 dBm/nm or -20 dBm/0.1 nm, 10 times less power because the spectral slice is 10 times narrower). Now if we connect a device under test that has 50 dB loss and measure it with the OSA, the power reaching the OSA in the 0.1-nm slice is now only -70 dBm.
The spectral stability of a broadband source is the variation over time of the power in a spectral slice. The total power stability is the variation over time of the total integrated power. It is relatively simple to design a source that has excellent total power stability; achieving spectral stability, however, is much more challenging. Care must be taken on how the specifications are written. Because of the nature of ASE sources and even light-emitting diode (LED) or superluminescent LED sources, there can be a transfer of energy from one wavelength to another. This transfer may not be measurable if we are simply looking at total power, but if we start to look at the stability of a small spectral slice, then significant variations in power can be observed.
When we are characterizing DWDM components, we generally want to have as high a resolution as possible (as small as possible a resolution bandwidth). If the power in each spectral slice is stable, there is no problem. If the power in the spectral slice is varying, measurement error can result.
Operating wavelength range
The operating wavelength range of the source is the range over which minimum detectable power density is achieved. Usually it is defined by the 3-dB bandwidth or the 20-dB bandwidth.
The light source must generate a spectrum broad enough to extend over the entire DWDM wavelength range. When such a source is connected to the entry port of a DWDM passive component, an accurate OSA can reveal the transmission characteristics of the component.
The quality of the measurement is as good as the quality of the tool. An OSA with 20-pm wavelength accuracy, resolution bandwidth less than 70 pm, and at least a 45-dB SNR ensures that the user sees the limits of the passive device, not of the OSA. High-end field OSAs now offer resolution bandwidth as low as 35 pm, greatly increasing the quality of the measurement.
A typical measurement sequence would be as follows: the source is turned on and left to stabilize; a reference measurement is performed (source connected directly to the OSA via the patch cords to be used later); a device under test is inserted; the measurement is performed; loss is calculated from the difference between the measurement and the reference
Use of this method has several advantages over the combination of tunable-laser source and power meter. First of all, a broadband source such as an ASE source can cover a wider wavelength range, which enables testing components such as coarse WDM devices, WDM couplers, switches, and attenuators. This is not always possible with a tunable-laser source. In addition, both the OSA and the ASE source have high power linearity, which is a great advantage when measuring spectral ripple of DWDM filters.
The high dynamic range and sensitivity of the OSA can result in a higher overall dynamic range than the tunable-laser source configuration. And finally, a power meter only displays power, whereas today's good field OSAs offer several integrated applications that automatically calculate the bandwidth, insertion loss, center wavelength, and ripple of the device under test (see Fig. 2).
The output spectrum simply needs to be compared with the spectrum of the source itself. If the reference measurements are made properly, and the OSA is properly calibrated, an accurate loss spectrum of the component is acquired. High-end field OSAs have an integrated function to automatically compare traces and qualify the transmittance accordingly.
If the components to test are multichannel devices, the setup should include a switch to enable automatic testing of all the channels. Some modular field platforms can include both OSA and switch at the same time, thus enabling testing multichannel components with a single field-portable unit. Such a platform can also include a multiwavelength meter for accurate calibration of the OSA.
FRANCIS AUDET is a product manager at EXFO, 465 Godin Ave., Quebec City, Quebec G1M3G7 Canada; e-mail: [email protected].