To accommodate rising traffic, designers of telecommunications components must provide higher bit rates and denser channel spacing. Optical power meters are a key element in testing these components. Despite the downturn in the telecommunications industry, traffic volume continues to grow. Network providers will need additional capacity, and equipment designers must continue to push system limits toward higher bit rates and denser channel spacing. The state-of-the-art commercial dense wavelength division multiplexing (DWDM) systems employ 80 to 160 channels with 50-GHz spacing. Systems with 25- and 12.5-GHz spacing are in development, and recent proceedings report experiments with up to 1000 channels on a 2.5-GHz grid. Very narrow channel spacing results in significant challenges to the manufacture and qualification of the components for these systems. Nevertheless, the testing systems must continue to satisfy the goal of component manufacturers: to produce these components economically at a high throughput while maintaining quality. Optical power meters are a key element in these systems for characterizing new and prototype products in the face of rising channel counts. Challenges in DWDM testing For DWDM components, many parameters must be tested using a few basic measurements. These parameters can be divided into two groups: • Loss-related, such as insertion loss, insertion loss uniformity, passband ripple, isolation and crosstalk. • Wavelength-related, such as operating wavelength, center wavelength, passband and bandwidth. Because the parameters of each group are related, they can be determined by a loss- vs.-wavelength measurement. The swept-wavelength method, which uses a tunable laser source and multiple broadband receivers such as optical power meters, is the method of choice, providing the highest possible wavelength resolution and accuracy as well as the highest loss accuracy. Measuring outputs simultaneously with multiple receivers greatly reduces the measurement time for multiple-output devices. A mode-hop-free, continuously tunable laser source sweeps at a constant velocity over a wavelength range. At defined wavelength intervals, the source triggers the power meters to synchronously record the power. A precise, built-in, real-time wavelength meter records the wavelength of the source at each trigger, and the measured wavelength and power values produce the transmission spectrum of the device under test. Both the loss- and wavelength-related parameters of the device are derived from this transmission spectrum. The specified performance of a given parameter cannot be better than the inherent uncertainty of the test system, so it is preferable to achieve the highest possible accuracy in the test procedure, within the constraints of testing time and equipment investment. The results are better specifications of the parameter in question, reduced test uncertainty and higher component yield. Dynamic range On the optical receiver side, several parameters are directly related to uncertainties in the loss domain. One of the most important is the available input power range. The input of the optical power meter usually sees the wavelength-dependent output power of the source modulated by the response of the device under test — for example, a DWDM filter. For this application, tunable lasers with a low source spontaneous emission output are preferred. Their output, combined with the receiver noise, limits the dynamic range of the system. A tunable laser with a low source spontaneous emission design drastically reduces the total emission power to a level better than 60 dB below the signal power. Tunable laser sources deliver an optical output power at their low source spontaneous emission output in the range of 0 to —13 dBm over the wavelength tuning range. The filter response consists of insertion loss in the passband and attenuation in the stop band. DWDM filters typically have an insertion loss in the passband of 5 to 10 dB and a stop-band attenuation of 30 to 40 dB. This results in a total dynamic range of 35 to 50 dB. (State-of-the-art filters employing thin-film technology, however, may have a dynamic range of 70 dB or more.) The combined source output power and dynamic ranges of the device mean that the optical receiver must be able to measure from 0 to 285 dBm. To cover such a large dynamic range, most optical power meters with a linear receiver design have several gain ranges. This type of design offers the best linearity over the whole dynamic range. Power linearity is one of the most important factors because all nonlinearities of the power meters directly contribute to the relative loss uncertainty of the measurement system. The best power meters feature specifications of 0.02 dB or less. Logarithmic or linear? Meters with a logarithmic receiver design provide a similar input power range in one gain range. At first glance, this appears to be advantageous, in terms of measurement speed, because it enables characterization in one wavelength sweep. Most linear power meters, in contrast, may need two sweeps and stitching of the result traces to cover the dynamic range. However, logarithmic power meters exhibit significantly larger nonlinearities of up to 0.5 dB. In addition, the bandwidth and phase response of the logarithmic receiver are not constant, creating distortions at the side lobes of the filter. In the results of swept-wavelength measurements at a sweep speed of 5 nm/s with both types of power meters for the same device under test, the characterization result of the logarithmic power meter shows strong oscillations at the side lobes, caused by nonlinear frequency response of the power meter at small signal levels (Figure 1). Figure 1. The choice of a logarithmic or linear optical power meter affects the results of swept-wavelength measurements. At a sweep speed of 5 nm/s, thelogarithmic meter displays strong oscillations at the side lobes. Courtesy of Agilent Technologies. The bandwidth of the analog receiver path in the power meter plays an important role in fast and accurate signal characterization, especially for DWDM devices with a channel spacing of 50 GHz or less (Figure 2). The filter transfer characteristic for one channel of a 2.5-GHz multiplexer is interpolated based on typical specification values such as 1- and 3-dB bandwidth and stop-band attenuation. Figure 2. The bandwidth of the analog receiver path in the power meter plays an important role in characterization. Here, the filter transfer characteristic for one channel of a 2.5-GHz multiplexer is interpolated based on typical specification values. In a logarithmic scale, the curve’s steepest slopes are at the transition of the side lobes to the stop band. In a logarithmic scale, the characteristic transfer curve has the steepest slopes at the transition of the side lobes to the stop band. These are the regions where power meters with logarithmic receivers exhibit very strong distortion. In a linear scale, the picture looks quite different (Figure 3). Here, the filter characteristics of Figure 2 are transferred to a linear scale. The steepest slopes are now around the 50 percent point of the side lobes, which is equivalent to the 23-dB points in the logarithmic scale. This means that the steepest slopes are in or around the passband of the filter. Figure 3. In a linear scale, the channel characteristics in Figure 2 look quite different. The steepest slopes are around the 50 percent point of the side lobes, in or near the passband of the filter. Sweep speed and distortion For an accurate characterization of parameters such as passband bandwidth, the power meters must perform signal recording with very low distortion. They can do this only if the analog bandwidth is high enough in relation to the sweep speed of the laser source. In other words, fast sweep speeds, which are essential for short characterization times and high testing throughput, require receivers with high analog bandwidth. On the other hand, a higher analog bandwidth increases the noise level of the power meter, which reduces the potential dynamic range. Today’s wideband tunable laser sources feature maximum sweep speeds of 100 nm/s, and speeds of up to 200 nm/s appear to be realistic in the near future. Filters with dense spacing can serve as benchmark devices to evaluate the requirements for the DWDM test systems of the next generation and to demonstrate the limits of today’s available test equipment. The maximum slope for these devices can be derived from the previously described interpolation of a 2.5-GHz filter characteristic. The linear scale gives a value of 45 percent per picometer, which equals approximately a 2-pm width in each side lobe. The minimum analog bandwidth of the power meter, BWpm, can be determined from the side-lobe width, Δλ, and the sweep speed, vsweep, with the formula: In the above equation, vsweep = 40 nm/s and Δλ = 2 pm, which yields a bandwidth >11.2 kHz. In Figures 4a and 4b, a DWDM device was characterized using Agilent Technologies’ 81632A and 81636B power meters, which support DWDM tests down to 2.5 GHz. The device under test in this case was an acetylene gas cell, which has similar slopes to a 2.5-GHz filter. Both measurements were executed using a test-and-measurement control software library that includes a correction algorithm for phase and amplitude distortions in sweep applications. The measurement with the conventional power meter, which has an analog bandwidth of less than 4 kHz, displays significant amplitude distortion at higher sweep speeds (Figure 4, left). The other measurement, in contrast, shows no distortion (Figure 4, right). Figure 4. The characterization of an acetylene gas cell reveals the effect of sweep speed on the measurement, using the —40-dBm Agilent 81632A optical power meter (left) and the —20-dBm 81636B (right). As discussed earlier, the dynamic range of the power meter is one of the most important performance parameters for swept-wavelength applications. Most DWDM devices require a dynamic range of up to 55 dB for their characterization. Power meters with a linear receiver design usually provide 45 dB in one gain range. This is insufficient to characterize most devices in one sweep. It is possible to sweep two to three times in different measurement ranges and to stitch the results in the application software. This provides the necessary dynamic range but extends the test time by at least the number of sweeps. A more sophisticated method is to stitch during the sweep. A high-speed dedicated stitching circuit sets the optimum gain range in the power meter and ensures that the readings are processed to the correct power values. Such a stitching circuit increases the dynamic range to more than 55 dB for one fixed-measurement range. The resulting advantage in speed is illustrated in Figure 5. The time of the sweep or sweeps is the same for any number of channels. The measurement time nevertheless increases with the number of channels because of the time required to transfer the greater data to the computer. Figure 5. Single-sweep optical power meters, such as the Agilent 81637B, stitch their results during the measurement sweep, resulting in faster sampling times (blue) than multiple-sweep power meters, such as the company’s 81632A, which stitch after the sweeps are complete (red). The span is 1520 to 1570 nm, the step is 5 pm, and the sweep speed is 5 nm/s. Power meters are a key element in the swept-wavelength test. Their main performance parameters — dynamic range and bandwidth — must be considered in conjunction with the tunable laser sources to obtain the best system performance. There is a great demand for products that offer high dynamic range and wide electrical bandwidth for high-speed, high-dynamic DWDM device characterization down to 2.5-GHz channel spacing. Meet the author Halmo Fischer is a research and development project leader and systems architect at Agilent Technologies’ Optical Communication Measurement Div. in Böblingen, Germany. He holds a Diplom-Ingenieur degree in communications engineering from the University of Stuttgart in Germany.