Optical networking within the realm of a transport network is defined as the ability to implement facility restoration and protection switching in the physical optical layer. From this definition comes the requirement for an optical network to conduct performance monitoring and channel routing in the physical optical layer.
Achieving the goal of a multichannel-path, reconfigurable, all-optical network requires the deployment of several enabling technologies, such as optical crossconnects (OXCs), optical add/drop multiplexers (OADMs), chromatic dispersion-equalization modules (DEMs), polarization mode-dispersion (PMD) compensators, and optical-performance monitors (OPMs).
Whether a network is based on point-to-point, mesh, or ring topologies, its optical elements can be strategically placed to perform such tasks as performance monitoring, performance optimization, and optical routing. Performance monitoring is accomplished by OPMs that are able to measure the characteristic optical parameters of a dense wavelength-division multiplexing (DWDM) signal and inform the network-management layer of the network's state of health. Performance optimization of the optical carriers is performed by DEMs and PMD-compensation modules that are able to reduce the harmful effects of chromatic dispersion and PMD in the route. Path routing at the add/drop node is attained by employing OADMs and OXCs to reconfigure network pathways to either balance capacity or, in the event of a severed link, initiate an optical-protection switch and avert a service interruption.
The ADM is the basic building block of fiber-optic network architectures that employ either unidirectional or bidirectional traffic configurations. Existing architectures based on the Synchronous Optical Network/Synchronous Digital Hierarchy (SONET/SDH) ADM use an electrical multiplexer that employs time-division multiplexing (TDM) to combine multiple inputs with varying subsidiary bit rates to the OC-N rate of the backbone network. At any add/drop site, only the data that need to be accessed are dropped, while the data to be inserted are time-division multiplexed with the rest of the through traffic and passed to the next node.
With the recent appearance of DWDM optical carriers in the same fiber, the add/drop function is now best managed in the optical layer. In this case, the required data are accessed by optically filtering a wavelength channel from the channels entering the node. The same optical carrier can be used as the add channel for data to be inserted into the network at that node.
In an optical network, the OADM is transparent to the express-through traffic. Each wavelength channel (at various OC-N rates) can be dropped or added without the need to multiplex/demultiplex TDM signals in the electrical layer. Another feature of wavelength add/drop is that wavelength-based services can now be offered, with the added economic advantage of being able to lease a wavelength channel instead of an entire fiber.
As the OADM site is a node where the DWDM optical carriers are amplified, generated, and routed, a majority of an optical network's functions occur at this node. A schematic of an add/drop site with optical-networking components such as the tunable OADM, channelized tunable DEM, and the OPM appears in Figure 1. These modules are compact, rugged, and wellsuited for embedded-system application. Furthermore, as each of these modules uses fiber Bragg gratings (FBGs) as a core component, they ultimately benefit from the advantages of in-fiber devices such as low insertion loss and minimal package integration.
The tunable OADM shown in Figure 1 can add or drop either a single wavelength channel or multiple channels. In the future, as more routing responsibility is shifted to the optical domain, the tunable OADM will need to undergo a smooth transition from a static configuration to one that is reconfigurable and eventually to one that enables dynamic operation. This evolution will allow the network-management layer to reconfigure the different wavelength paths to either avert a service interruption caused by a faulty span or simply balance the network's utilization to improve transport efficiency.
Current OADM schemes are based on thin-film dielectric filters, acousto-optic tunable filters, arrayed waveguide gratings, or FBGs. Dielectric filters have been actively deployed as OADMs for systems with channel spacing of 100 GHz and above. For 50-GHz spacing, dielectric filters are hard to produce, as the requirements for slope and center of the passband are demanding targets to meet. Acousto-optic filters have the advantage of wide wavelength tunability but lack the appropriate passband-
filter characteristics. Arrayed wave guide gratings are cost-effective for high-channel-count systems, but the insertion loss of the device still remains high. Much like the acousto-optic filters, arrayed waveguides lack the appropriate passband-filter characteristics.
The FBG is an in-fiber grating written in the core of the fiber. By virtue of being able to change the design parameters such as the length, index modulation, and optical pitch of the fiber grating, a large variety of filters can be made. They range from dispersion-com pen sating gratings (DCGs) to DWDM gratings for channels with 50-GHz spacing. FBGs can be made wavelength-tunable by mechanically loading the fiber in compression or tension. Temperature can also be used to tune the wavelength of the FBG, but the practical range is small and the tuning speed is slow. Mechanical loading methods have shown wavelength tunability up to 48 nm. For practical purposes, a range equivalent to the erbium-doped fiber amplifier (EDFA) band will satisfy the current requirements of wavelength-selectable OADMs. Larger tuning ranges raise concerns of fiber reliability.
The wavelength-tunable FBG device offers many advantages when incorporated in an OADM. In particular, the low insertion loss and a passband shape with the ability to accommodate 50-GHz channel spacing with the required channel isolation are the most compelling features. For flexible bidirectional switching, the FBG must be attached to a low-stiffness structural member. Mechanical loading of the structural member with a solenoid-based actuation system offers fast switching to accommodate the 50-msec sonet/sdh protection-switching requirements. To ensure the required repeatability specifications, the loading must be well within the elastic deformation regime of the structural member. A necessary environmental requirement is the passive temperature compensation of the FBG once it is statically configured. For deployment, all optical-networking modules should be qualified for temperature and humidity stability and shock and vibration performance (as per Telcordia Technologies GR-63).
Figure 2 shows the wavelength-tuning results for a 50-GHz DWDM FBG with an achievable dynamic range of 600 GHz. To allow total flexibility in accessing any channel within the EDFA band, eight concatenated tunable FBG devices would be required to cover this entire band. These eight devices can be placed between a pair of 3-port circulators to compose a tunable OADM module with four functional ports: in, through, add, and drop. In this device scheme, the insertion loss per channel for the through, add, and drop ports are typically not more than 2 dB. Loss equalization is achieved because all wavelength channels go through the same number of circulator paths with almost complete transparency outside the passband of the FBG. Wavelength dependencies such as the passband flatness (<0.05 dB), polarization-dependent loss (<0.2 dB), and dispersion are within acceptable criteria. Furthermore, the PMD of the tunable OADM is low, as the dispersion of the DWDM gratings are controllably small within the passband.
OPMs can be employed at numerous locations along a network. For example, an OPM located at the receiver line terminal is important for network commissioning and long-term monitoring. One located at the transmitter end can perform the added task of wavelength locking to an accuracy of +20 pm. Between the line terminals, an OPM can be employed to measure either the successful reconfiguration of a tunable OADM or the imbalances in the composite DWDM signal caused by separate channels having traveled along different paths in the network.
At the add/drop node, embedded OPMs are responsible for the optical-layer monitoring of the wavelength channels. In particular, they measure the important optical parameters of the individual DWDM channels, such as wavelength, power, and optical signal-to-noise ratio (SNR) (see Fig. 3). From these measurements, the number of channels, the channel spacing, the relative channel powers, and the amplifier gain and gain tilt (which is the flatness of an optical-line amplifier) can be assessed. Furthermore, to augment end-of-line bit-error rate (BER) monitoring, SNR measurement by an OPM can be used to infer the BER of a channel at mid-span.
Optical-performance monitors are designed to operate as embedded optical-spectrum analyzers (OSA). But unlike their benchtop counterparts, they must be compact and self-calibrating, have a higher degree of reliability, operate under a broader range of environmental conditions, and be cost-effective for wide deployment. Furthermore, the OPM must have sufficient accuracy and speed of measurement to enable it to report timely, useful information to the management layer. Accuracy and speed specifications can be dependent on the OPM's location in the network.
In general, to measure the spectrum of a source, either a dispersive optical element or a tunable narrow-bandpass filter must be employed. Several competing technologies incorporating one of these two basic designs have emerged that enable OPM manufacturers to meet the stringent requirements outlined above. These solutions include designs based on bulk gratings as per traditional OSAs, blazed gratings coupled with linear detector arrays, narrowband Fabry-Perot (FP) tunable filters, and narrowband tunable fiber Bragg gratings.
Although each solution has its share of weaknesses, the FP filter has garnered the most interest, and devices employing FP filters are now offered by a number of suppliers. The strength of the FP design is the very narrow passband that can be achieved by using high-finesse filters. This sharp Airy function is required to effectively discern the noise level between narrowly spaced channels. The currently available products work well for channels with a spacing of 100 GHz or more. However, considerable signal processing is required for systems with 50-GHz spacing.
At an add/drop node, an OPM is required to measure wavelength to an accuracy suitable for correct channel identification, which in practice means an accuracy to within +100 pm. Conversely, high accuracy for power and SNR measurements at an OADM node is required to properly diagnose noncatastrophic failures (e.g., channel degradation due to a degraded amplifier). Absolute power accuracy of +0.5 dB is the current standard. Achieving greater accuracy will depend on improvements in calibrating references, since benchtop OSAs and wavelength meters are typically only accurate to within +0.4 dB themselves. OPMs are generally calibrated with continuous-wave sources so that the measured power of any particular channel will depend on the modulation scheme being employed. Fortunately, it is the relative power between adjacent channels that is important, and OPMs are well-suited to deliver accurate relative power measurements as systematic errors cancel.
A final important parameter for power measurements is the dynamic range. OPMs employed with a 5% tap need to be able to measure carrier powers ranging from -55 dBm to +5 dBm. The response time of an embedded OPM depends, as with a traditional OSA, on the strength of the carrier signal; however, measurement times on the order of 1 Hz are typical for signals with higher powers.
SNR is typically specified to be accurate within a certain value of an OSA set to a resolution bandwidth of 100 pm. Commercially available OPMs typically specify an accuracy within 1 dB of an OSA when the SNR varies from 15 dB (which is the minimum received SNR) to 35 dB (which is the SNR at launch). For 50-GHz-spaced channels with an SNR of 35 dB, the raw SNR measurements from an OPM are typically on the order of 15 dB. The discrepancy is caused by the non-ideal width of the FP filter; therefore, a correction algorithm is required to match the performance of the OSA. Alternatively, tunable FBGs are a promising technology, as the bandpass response of an FBG can closely approximate the slit function that is found in an OSA. Fiber gratings can be made with a sub-angstrom bandpass and have been shown to be tunable over a large wavelength range.
As planners implement the reconfigurability of wavelength paths in the network, there are significant factors to consider that affect the performance of the system. The immediate concern is the equalization of the DWDM channels after an add/drop node. Optical characteristics such as the dispersion, power, and SNR of the individual channels have to be adjusted to accommodate the new traffic routes.
There are several ways to equalize the power and SNR. The simplest is to adjust either the channel-dedicated variable optical attenuator or the gain of the optical-line amplifier. Another, more-complicated method of equalization is to select alternate routing. The dispersion, however, can only be equalized on a per-channel basis and requires a remotely controlled, tunable-dispersion-based device that will be capable of on-the-fly dispersion management of reconfigured traffic. In effect, such a module will allow the optical channels to remain dispersion-equalized even though the wavelength path has been switched.
Narrowband in-fiber DCGs can be made dispersion-tunable for adjust able equalization. They are well-suited for embedded applications as they can be fit into a compact package on a network card. Furthermore, the static and reconfigurable nature of tunable DCGs makes them an ideal device for optical networking.
The DCGs used in one tunable DEM were fabricated using 10-cm-long holographically patterned phase masks, a standard procedure for producing high-quality FBGs. The dispersion of the DCG in the module is mechanically tuned by controlling the linear chirp in the grating. This control is exercised via the strain-imposition method, in which a precise linear strain distribution is applied to the DCG such that a linear chirp is induced with no wavelength shift. The deformation can be imposed by using a stepper motor to continuously tune the DCG with a resolution of approximately 17 ps/nm.
Figure 4 shows group-delay plots for a tunable DEM with five dispersion-set values ranging from -1270 to -420 ps/nm. The series of group-delay responses were measured over the 1-dB bandwidth of the DCG with a standard test set that employs the modulation phase-shift technique. The group-delay ripple at 640 psec/nm shows a nonlinearity excursion of less than +10 psec in amplitude, which is expected to have a minimal impact on the variation of the BER at a system level. Furthermore, the insertion loss of the entire device amounts to less than 1.8 dB at a setting of -1300 ps/nm, making DCG-based DEMs a very attractive optical-networking component.
In summary, the OPM, tunable OADM and tunable DEM modules perform the critical network tasks of performance monitoring, configurable multiplexing, and dispersion compensation. These technologically ready and network deployable modules are well-suited for embedded application. Together they offer great potential as core elements of the physical layer of future networks. With the need to increase bandwidth cost-effectively, it is clear that advanced optical components, which offer adjustability in their performance and versatility in their application, are absolute key requirements. u
Vislosky, T.W., "System Issues Associated with Wavelength Add/Drop," NFOEC Technical Proc., Vol. II, pp. 383-391, Sept. 13-17, 1998, Orlando, FL.
Myo Ohn is market and product manager, and Keith Beckley is a product manager at E-TEK ElectroPhotonics Solutions (Markham, ON, Canada).