High-bandwidth networks bring new monitoring requirements
To manage dense wavelength-division multiplexing networks, designers can choose from several grating/detector- and interferometer-based technologies.
Sami T. Hendow
Dense wavelength-division multiplexing (DWDM) is the accepted solution for increasing telecommunications network capacity while controlling overall system cost. Currently, several manufacturers offer multiple-wavelength DWDM systems using channel rates up to 10 Gbit/s. Some of these systems are scalable to 400 Gbit/s at 0.4-nm channel spacing. Wavelength allocations for the systems being deployed today all follow the International Telecommunications Union (ITU) recommendation of 193.1 THz 𫏌 GHz. Currently, manufacturers offer a variety of channel spacings such as 200 GHz (1.6 nm), 100 GHz (0.8 nm), and 50 GHz (0.4 nm).
Dense wavelength-division multiplexing is enabled by the wide bandwidth of the optical fiber, as well as the bandwidth of erbium-doped fiber amplifiers. Amplifier bandwidth is usually the flat region of 1540 to 1560 nm, or about 1525 to 1565 nm for gain-flattened amplifiers. Increasing demand for bandwidth has further stimulated research on extended-bandwidth amplifiers, which are reported to have bandwidths of approximately 1530 to 1610 nm.
Monitoring required
Most of the installed optical-fiber networks assume a point-to-point link between sites. This architecture is being rapidly expanded to include add/drop capability for channel wavelengths using switches and routers that control channel destination. Such desired network flexibility requires monitoring for operation and management. For example, a change of the power of an added channel can degrade the signal-to-noise ratio (SNR) of other channels or, alternately, a rerouted wavelength may not have the needed SNR to carry traffic if injected into routes that do not have ample safety margins.
Furthermore, monitoring the status of DWDM channels has become a requirement to minimize network down time, as well as to anticipate problems such as aging or drift of individual transmitters. Also, environmental conditions or damage to the fiber cable can degrade some or all transmissions. This variety of events makes it desirable that network managers monitor all network operating conditions simultaneously. Knowing the location and source of a fault goes a long way toward minimizing repair time or the number of affected calls (see Fig. 1).
The most significant parameters of DWDM network operation that require monitoring are channel power, signal-to-noise ratio (SNR), and channel wavelength. Channel power and SNR are affected by an accumulation of factors such as insertion loss, polarization-dependent loss, and amplifier gain of the various in-line components in the network. Channel wavelength is driven by the transmitter`s wavelength. If it drifts beyond its specifications, which are very tight for 50-GHz channel spacing, it contributes to its failure, as well as the failure of neighboring channels.
Analyzing the spectrum
A variety of optical spectrum analyzers are available to determine channel wavelength and its stability during network operation. Five kinds of optical spectrum analyzers (OSA) are described here, each with its own merits and shortcomings (see Fig. 2).
Rotating grating/fixed detector. This OSA configuration has a rotating grating, which allows for wide spectral range (600 to 1700 nm). The approach also accommodates a double pass over the grating, which gives the signal dynamic range of a double monochromator (-65 dB at 1550 nm) with the sensitivity of a single monochromator (-90 dBm at 1550 nm), as well as polarization insensitivity. An additional internal absorption cell helps in wavelength calibration. However, moving parts, as in the direct-drive motor system for grating tuning, generally make the mechanism sensitive to vibrations and shock. The majority of units in laboratories today use this configuration.
Fixed grating/scanned detector. The detector in this OSA configuration is scanned against a stationary grating. This arrangement reduces the number of moving parts, making the OSA less prone to shock and motion; however, the benefits come at the expense of a reduced wavelength range of 1525 to 1570 nm. The moving detector also slows the data acquisition and integration cycles. Typical resolution bandwidth of 0.1 to 0.5 nm, amplitude measurement accuracy of <0.8 dB, and small size make it convenient for characterizing WDM networks.
Scanning Michelson wavelength meter. By counting the number of fringes as one arm of a Michelson interferometer is extended, one can measure a channel wavelength to a very high degree of accuracy. In the case of multiple wavelengths, counting fringes is insufficient to extract their spectral profiles. However, by measuring the amplitude of these fringes as the interferom eter arm is extended, one can calculate the full spectrum of the input by performing a fast Fourier transform (FFT) calculation of the amplitudes.
This approach has the advantage of wide wavelength range (700 to 1650 nm) and wavelength accuracy of 10-2 to 10-4 nm for a single input wavelength, as well as 0.16-nm resolvable separation between input lines and power-measuring accuracy of less than 1 dB for multiple-input wavelengths. However, the response time of the instrument is reduced due to the combination of a scanning mechanism, integration time, and FFT analysis.
Scanning Fabry-Perot interferometer. A scanning interferometer consists of two parallel and closely spaced (30 to 50 µm) mirrors, separated by a piezoelectric transducer (PZT) spacer. By applying a voltage to the PZT, the Fabry-Perot interferometer mirror separation changes, allowing light to be transmitted through it if mirror spacing is a multiple of half of the wavelength of the input. However, PZTs are inherently prone to drift. To account for drift, Fabry-Perot interferometers require an additional independent reference for wavelength calibration, such as an internal absorption cell. Alternately, an external capacitor and a capacitance bridge system servo can be added for mirror spacing measurement and stabilization.
This approach, in combination with a high-resolution Fabry-Perot interferometer, can produce a compact, solid-state, and board-mountable device. The spectral transmission characteristics of a Fabry-Perot interferometer, however, limit its ability to reject wavelengths adjacent to the peak, which in turn limits the dynamic range and SNR measurements. To improve isolation better than 25 dB at 0.8 nm (at spectral range of 40 nm), the Fabry-Perot interferom eter requires either finesse values greater than 350 (where finesse equals spectral range/resolution) or a multipass configuration to improve rejection. The wavelength spectrum is developed by scanning the mirrors and averaging over the spectral range.
Fixed grating and linear detector array. This OSA configuration, the most recent technology for network-servicing products, combines a grating and linear detector array. It is highly suitable as a deployable instrument throughout the network because it has no moving parts that might drift with shock or temperature cycling or simply wear out with age, thereby requiring periodic calibration. Its most significant attribute is the ability to process optical channels in parallel without a scanning mechanism, which considerably speeds up data acquisition and alarm reporting. In contrast, the four technologies discussed above process channels serially with an internal wavelength scan.
The internal configuration is composed of an input fiber, mirrors, grating, and an InGaAs linear detector array. The design is configured to produce a dispersion of about 1530 to 1560 nm across the array. Therefore, each channel is allocated a certain number of detector pixels, where optical power is measured simultaneously for all channels. In addition, the location of the peaks for each channel also defines accurately the wavelength of each channel. The signal-to-noise ratio is simply the ratio of power intensities measured by different pixels.
The disadvantage of this instrument is the high initial cost of the detector arrays. However, the cost is expected to drop as volume increases with deployment. Such an instrument also is suitable as a network-service device for debugging and installation.
Alternate monitoring techniques
Another approach used to monitor channel power in a DWDM network is to modulate the carrier for each channel at a specific coded frequency. Channel power can be inferred by monitoring the amplitude of this filtered frequency. This method integrates the monitoring function into the multiplexers and transmitters for each channel and is effective for amplitude monitoring of individual channels.
Channel wavelength, however, remains unmonitored. Consequently, if a transmitter laser wavelength drifts, it would be undetected unless additional means are implemented for analyzing the signal-to-noise ratio for each channel. As channel spacing decreases with increases in system capacity, it becomes imperative to find problems before they occur, particularly if the faulty transmitter laser has the potential to drift across neighboring channels, driving those channels out of commission, too.
The availability of remote channel monitoring opens the door for other network applications. For example, by monitoring the amplitude of all channels and reporting it to the transmitter end, one can actively flatten the received spectrum of all channels simultaneously by pre-emphasis of the transmitted amplitudes. This closed-loop control of amplifier flattening by pre-emphasis may cause degradation of channel signal-to-noise ratios and thus be undesirable. However, it may still be advantageous in situations where overall system integrity must be maintained even if a few channels arbitrarily fail.
In addition, the speed of data acquisition and display of information using the parallel channel processing of a detector array may be highly advantageous in viewing transient effects. For example, it can reveal the effects of changes in polarization-dependent losses or backscatter or of instabilities previously undetected because of the slow response of spectrum analyzers based on scanning-and-averaging mechanisms. o
FIGURE 1. Optical network monitors placed at varying locations in a DWDM system make it possible to continuously monitor the status of all channels. Site controller at each monitor location integrates operation and alarm reports into the supervisory channel, typically at 1510 nm (not shown).
FIGURE 2. A variety of spectrum analyzer configurations are available. Rotating grating/fixed detector combination with a double pass over the grating allows for both wide spectral and dynamic ranges with polarization independence (a); fixed grating/scanned detector combination generates smaller package with reduced spectral range (b); scanned Michelson wavelength meter offers high spectral accuracy for single-wavelength input; an FFT is added to analyze multiwavelength transmissions (c); scanned Fabry-Perot interferometer with capacitive feedback for mirror spacing stabilization is used to compensate for PZT drift (d); fixed grating and linear detector array combination offer similar optical performance as above, with the added advantage of parallel signal processing; no moving parts make it ideal for long-term field installations.