Study of the performance of a transparent and ... - Semantic Scholar

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 6, JUNE 2002

937

Study of the Performance of a Transparent and Reconfigurable Metropolitan Area Network Nicholas Madamopoulos, Member, IEEE, Member, OSA, D. Clint Friedman, Ioannis Tomkos, Member, IEEE, and Aleksandra Boskovic, Member, OSA

Abstract—The transport performance of an optically transparent regional-size ring network testbed with circumference of 280 km, based on metro-area optimized optical layer components and fiber, is demonstrated under dynamic traffic conditions. For the longest transmission path, excellent transmission performance is achieved using cost-effective directly modulated signals. Network reconfigurability is achieved using add–drop modules that are commercially available as of this writing. We show that the dynamic nature of the network does not affect the system performance. In particular, we show that electronic gain control of erbium-doped amplifiers is capable of managing switching transients in amplified metro-scale networks. Index Terms—Amplifier transients, metropolitan area networks, optical network performance.

I. INTRODUCTION

T

HE GROWTH of telecommunications, due mainly to the evolutionary growth of the Internet has greatly increased the bandwidth demands in optical networks. Both existing and future applications lead to metropolitan area networks that must support quick provisioning and frequent reconfiguration to meet customers needs. This very diverse environment is becoming harder to be served by traditional opaque networks based on optoelectronic (OE) and electrooptic (EO) conversions. Transparent dense wavelength division multiplexing (DWDM) can eliminate the expensive transporters needed for the OEO conversions in an opaque approach [1], [2]. Nevertheless, the infrastructure of a metropolitan area network is shared among fewer people than in the case of long haul-networks. Hence it is important for new more cost-effective solutions, that provide network configurability and scalability, to gain greater acceptance for transparent optical metro network to become a reality. Optical switching technologies and configurable optical add–drop multiplexers (OADMs) have the ability to serve the dynamic nature of data traffic in a metropolitan area network. Since no OEO conversions are required, support of multiple traffic formats is possible. Architectural studies, experimental proof-of-principle demonstrations, and field trials for metro-applications have been previously presented [3]–[9]. Studies have shown that the typical maximum connection length in metropolitan area networks is 280 km [10], [11]. Recent developments in optical network components and fiber have shown that such connection distances are feasible with proper Manuscript received November 6, 2001; revised February 21, 2002. The authors are with the Applied Research, Photonic Research and Test Center, Corning Incorporated, Somerset, NJ 08873 USA (e-mail: [email protected]). Publisher Item Identifier S 0733-8724(02)05386-0.

network design and engineering, and when metropolitan area network-optimized components are used [12]. The main impairments that limit the size of a metro area network are component insertion loss, optical power divergence of the different channels, noise accumulation, fiber and component chromatic dispersion, polarization dependent loss (PDL), polarization mode dispersion, laser frequency chirp, linear crosstalk, filter concatenation and fiber nonlinearities [13]. A signal in a metropolitan area network may traverse many different optical elements that provide different functionality and have a different degree of the aforementioned impairments. It is important that network engineering is used to optimize the performance of a network as it pertains to specific application spaces [13]. Amplifier power transients have to be considered when engineering transparent metro area networks. In this paper, we present the transport performance results under dynamic traffic conditions of a metropolitan area ring network testbed that uses different technologies for the OADM nodes, with 18 channels modulated at 2.5 Gb/s [14]. The OADM nodes offer different degree of add–drop functionality in the network. We show how the proper choice of optical components and fiber can provide the basis for optimized network performance. In this network testbed, direct modulated lasers (DMLs) are used as transmitters because they offer a cost-effective solution. DMLs may have significant frequency chirp and their output power waveform may not be an exact replica of the modulation current. First, the frequency chirp interacts with the fiber dispersion [15], [16]. Second, the frequency chirp broadens the optical spectrum, which leads to filter concatenation penalty [17]. Nevertheless, by choosing Corning MetroCor fiber as the propagation medium we can demonstrate high system performance [18]. Metro-area network optimized erbium-doped amplifiers (EDFAs) with gain control are also used. In particular, we show that electronic stabilization of EDFAs is capable of managing switching transients in amplified optical metro-networks. Section II presents the experimental setup and describes the different node designs. We show how different add–drop functionality is obtained using different types of OADM nodes. In Section III, the performance of the network is presented, in a static add–drop configuration. Excellent system performance is obtained for the longest path in the network testbed. The effect of the dynamic add–drop of channels in the system performance is discussed in Section IV. Dynamic gain controlled amplifiers are shown to be suitable to support excellent system performance in the presence of traffic disturbances that affect either the total number of wavelengths through the system or the total optical power.

0733-8724/02$17.00 © 2002 IEEE

938

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 6, JUNE 2002

Fig. 1. A typical view of a metropolitan area network where interconnection among the access rings is accomplished through a feeder ring.

II. METROPOLITAN AREA RING EXPERIMENTAL SETUP A possible metropolitan area network configuration is shown in Fig. 1. In this network, a feeder ring is used to transparently deliver signals to local nodes or access rings. The add–drop nodes may either route/switch signals in a “per wavelength” or at a “per band” fashion. It has been reported that wavelength banding may be of increased importance in metro area networks, where band filters may be used to transport several channels between nodes [19], [20]. Hence, banding can eventually reduce insertion losses and limit impairments due to filter concatenation. Based on the concept of Fig. 1, we built an experimental setup of a metropolitan network, as shown in Fig. 2, consisting of three reconfigurable OADM nodes. These three nodes are based on different technologies and will be described later. They also provide different degrees of add–drop functionality in the network. The three nodes are connected with each other using negative dispersion MetroCor fiber. The span lengths between node 1 and node 2, node 2 and node 3, and node 3 and node 4, are 100, 100, and 80 km long, respectively. The wavelength plan consists of 18 channels grouped in three bands. The first and second bands have seven channels each at 100-GHz wavelength spacing, with the first channel at 1538.18 nm and at 1547.51 nm, respectively. The third band has four channels at 100 GHz spacing with the first channel at 1557.36 nm. The first node in our system is a thin-film filter-based, manually reconfigurable OADM that has both channel and band add–drop capability (Fig. 3). More specifically, node 1 can perform per channel add–drop for band 1 (i.e., individual access to all of channels 1–7) and band add–drop for bands 3 and 4. Fig. 3 shows optional patch-cord connections for pass through configuration. If no patch cords are used, then the traffic is dropped at this site. Note also that optical switches can be used in the node to upgrade it to a remotely reconfigurable design. The second node is a low-channel count (e.g., four channel) remotely reconfigurable OADM node that is based on thin-film

filter technology with optomechanical activation. This OADM can independently access four predefined channels from the incoming traffic, either letting them through or dropping them to single wavelength drop-ports. The specific device used in this experiment allows for access of every other channel (i.e., channel separation of 200 GHz), as shown in Fig. 4. Fig. 4 shows seven incoming channels. The module is designed to access channels 1, 3, 5, and 7, while all other channels are passed through the node (express channels). In the specific state shown in Fig. 4, channel 7 is completely dropped, channel 5 is dropped and a new signal is added at the same nominal wavelength, while channels 1 and 2 pass through the node. The third node is a high-channel count remotely reconfigurable add–drop node (e.g., Corning PurePath WSS) [21]. The WSS is based on liquid crystal technology and has the capability of switching individually up to 80 channels spaced at 50 GHz, either letting them through or dropping them to a WDM drop-port. Hence, this node can be used to either drop–add any of the individual channels or bands of channels as shown in Fig. 5. In the example of Fig. 5, the module is set to pass through band 1 as a whole. Band 2 is dropped while a new set of signals at the wavelengths of band 2 is added. In band 3, individual access to wavelengths is used. That is channel 15 and 17 pass through and channels 16 and 18 are dropped and new signals at the same nominal wavelengths are added. In our experiment, we do not fully utilize the entire add–drop bandwidth of the WSS since only 18 wavelengths at 100-GHz spacing are used in the testbed. All unused channels at the WSS are set in pass through mode. Amplifiers tailored for metro area networks were used. The amplifiers used were PureGain 2400C for the first, third, and forth amplification stage and PureGain 2500C for the second amplification stage. Both amplifier types are variable gain metro amplifiers, where the gain can be adjusted to specific applications. In particular, the PureGain 2500C accomplishes this

MADAMOPOULOS et al.: STUDY OF THE PERFORMANCE OF A METROPOLITAN AREA NETWORK

939

Fig. 2. The experimental setup of the reconfigurable metropolitan ring built in the lab.

Fig. 3. The first node is a manually reconfigurable node based on thin-film technology.

without impacting the overall flatness. The amplifiers also have gain control that provides optimum gain stability for the reconfigurable system. Hence, added/dropped channels do not affect the performance of the remaining channels. These amplifiers have an adjustable gain of 10 to 25 dB over 33 channels in the C-band, with maximum output power of 13 and 17 dBm for the PureGain 2400C and 2500C, respectively. The noise figure is dB for gains of 20–25 dB. III. OPTICAL PERFORMANCE OF THE RING NETWORK In this section, we present the transport performance results obtained for the network under study. We present detailed results for two of the channels used in the network (channel 17 and 18) and overall signal performance for the rest of the channels. In these results we refer to the signal -factor evolution and optical signal-to-noise ratio (OSNR) evolution through the network, after the different OADM nodes and amplification stages.

All channels, except channel 17, originate from transient chirp dominated transmitters [18]. Channel 17 originates from an adiabatic chirp dominated transmitter [18]. We accomplish this in order to evaluate the system performance under different transmitter types. All signals are added in the network at Node 1. To account for the worst case scenario, we allowed the signals to propagate through the entire network and be dropped at Node 1. Hence, propagation through 280 km of fiber and 3 OADM nodes is obtained. The transmitters were modulated at 2.5 Gb/s with an extinction ratio of 8.2 dB [22], [23]. The signals were detected using a 2.5 Gb/s APD receiver and the optical fiber loss was padded at 0.23 dB/km. The launched power at the first fiber span was 0 dBm per 6 dBm/channel and at the channel, at the second fiber span 11 dBm/channel. The lower powers at the second last span and third spans ( 6 and 11 dBm, respectively) are due to the insertion loss of the OADM and the fact that the power into the amplifiers, preceding the node, was low. Hence, the maximum output power after the OADM nodes gives a low launch power into the fiber spans. These low powers eventually lead to a low input power at the amplifiers, which in turn limits the OSNR performance of the system. The OSNR performance can be improved significantly with the use of additional amplifiers in the system; nevertheless, we had only four amplifiers available for our experimental demonstration. Furthermore, this allowed us to study the network testbed under OSNR limited performance. The OSNR evolution for several channels through our network is shown in Fig. 6. The OSNR of the received signals after going through the longest path in the network was 20 dB (measured at 0.1 nm resolution bandwidth). Fig. 7 shows the evolution of for channels 17 and 18 due to all limiting effects (e.g., fiber dispersion, OSNR, filter concatenation). The results are presented notation, where is the linear . It is using the 10 evident, from the results of Fig. 7, that the signal performance is dominated by OSNR degradation while other impairments, and

940

Fig. 4.

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 6, JUNE 2002

The second node, wavelength access module (WAM) node, is based on thin film technology and is a reconfigurable node.

Fig. 5. The third node, wavelength selective switch (WSS) node, is based on liquid crystal technology and is a reconfigurable node with either channel or band add–drop functionality.

Fig. 6. The OSNR evolution through the system.

Fig. 7.

The

Q-factor evolution for channels 17 and 18 through the system.

in particular dispersion, do not significantly affect the system performance. Fig. 8 shows the factor at the end of the system for all the channels as well as their back-to-back performance.

Fig. 8.

Q-factor at the end of the system for all the channels.

The worst factor was observed for channel 5, which also had the worst back-to-back performance. From the results, we see that the combination of positive chirp of the transmitter and negative dispersion of the fiber (MetroCor fiber) allows for increased performance in our system. This interaction can be seen from the optical eye diagrams of Fig. 9(b) that show the optical eye for channel 18 through the network. Fig. 9(a) shows the eyes of channel 17, which is an adiabatic chirp-dominated transmitter, and the spreading of the pulse. We also evaluated the performance of the network for detuned transmitters to account for possible drifts over the lifetime of the system band potentially higher crosstalk and filter concatenation penalties. For a nm no performance degradation was wavelength offset off observed (less than 0.1 dB -factor penalty), as expected [17]. IV. TRANSIENTS IN THE RECONFIGURABLE OPTICAL NETWORK As mentioned earlier, the ever-changing traffic leads to the dynamic add–drop of channels and hence to the dynamic na-

MADAMOPOULOS et al.: STUDY OF THE PERFORMANCE OF A METROPOLITAN AREA NETWORK

941

Fig. 9. The evolution of the optical eyes for channel 17 and 18 through the system showing the interaction of the transmitter chirp and the negative dispersion of the fiber.

ture of reconfigurable networks. Furthermore, protection and restoration mechanisms in an optical network may force the abrupt termination of some channels/traffic. This sudden change in the number of channels changes the total optical power in the network. Hence, a varying optical power is impinging on the amplifiers, which must follow these changes. In amplified optical networks, the amplifiers are usually set to operate near saturation. Since the total output power of a saturated EDFA is constant (or almost constant) and independent of the number of channels passing through it, the gain experienced by each channel will depend on the total number of channels present. Therefore, a sudden change in the number of channels can lead to transients in the gain at different wavelengths, via cross gain saturation in the amplifier. This cross gain saturation will lead to power transients in the surviving channels. This means that when channels are dropped/added, the surviving channels may experience a large upward/downward transient spike in power that may last up to a millisecond. The power excursions may result in OSNR changes, due to inversion-level changes in the EDFAs, and may probe nonlinear effects in the fiber. Such effects may lead to a significant degradation of the bit-error-rate (BER) performance during the transients [24]–[27]. Changes in the inversion level of the amplifier may be beneficial for the system performance in some cases. If optical noise dominates the system, then the BER will be benefited from a power overshoot, as the OSNR will be improved and more power will be impinging on the receiver during the transient. However, if optical noise is not the dominant impairment, then other degrading effects such as increased crosstalk and nonlinearities during a power transient overshoot, starts to become a more important performance degradation effect. Moreover, a permanent offset in the amplifier gain or an unwanted power offset may be exhibited. Therefore, in a network where cascades of amplifiers are used, this transient effect and the BER degradation will be intensified. Hence, it is necessary to keep the amplifier gain at a constant level, independent of the number of wavelengths present in the network. This can be accomplished via EDFA gain control techniques. One of these techniques is based on keeping a probe (tone) signal at the output of the amplifier at a constant power level, by adjusting the pump power [28], [29]. Another approach is to control the ratio of the remnant pump signal and the launched value of the pump signal [30], [31]. All-optical gain-clamped amplifiers [25], [32] are another, not yet practical

technique, in which a lasing action at a wavelength outside of the signal band is used to control the amplifier gain. Some of the automatic gain control schemes are based on control of total power transients, by trying to keep the output signal power at a prespecified level with respect to the total input power [33], [34]. In this paper, we are using EDFAs that use monitoring of the total optical power at few points of the amplifier to control gain and output optical power [35]. In an ideal case, the gain control mechanism should sense changes in the total input or output power of the EDFA and adjust its gain such that a constant and flat gain is obtained for all channels independent of the number of channels impinging into the EDFA. If this is fast enough, then no power transients at the surviving/remaining channels should be observed. Nevertheless, it is very hard to accomplish an extremely fast control mechanism and hence power transients are observed. The effectiveness of the gain control mechanism is then related to the degree of power transient suppression it can offer as well as its time response in correcting the power transients. In this study, we refer to the channels that remain in the optical path after either protection/restoration mechanisms have been established or specific traffic (several channels) has been terminated/dropped at an OADM node as surviving channels. We examine the effect of the dynamic nature of the network on the performance of the surviving channels for loss of signal of 0.2–5 ms in duration. This disturbance in the total number of channels can be considered as the case of a protection/restoration event, where a portion of the traffic is lost for some amount of time. The amplifiers used in our system have dynamic gain control that provides optimum gain stability for the reconfigurable system with response times as fast as 1 ms. The amplifier control is accomplished via monitoring a portion of the input and output optical power using photodiodes. These received signals are then used to calculate the necessary control signals for the pumps and/or the variable optical attenuator in the amplifier via electrical signal processing and feedback control loops [35]. To study the transients in our network, we modulate, in an on/off mode, channels 1 through 17 using an acoustooptic switch (AOS). The AOS is placed at the output of the first node just before the amplification stage, as shown in Fig. 10. A 3-dB coupler is used to combine channel 18 to the rest of the WDM signals. The driving signal of the AOS is provided combining the carrier radio frequency (RF) of 35 MHz with a square wave

942

Fig. 10. The experimental set-up for accomplishing on/off modulation of the 17 channels for the transient studies.

Fig. 11. (a) The response time of the dropped/lost channel 17 with disturbance duration 2 ms. (b) The response of channel 18 showing the transients suffered by the sudden drop of 17 channels. (Repetition rate: 100 Hz.)

signal through a mixer. Two repetition rates are chosen for our experiments, 100 and 500 Hz, giving a period of 10 and 2 ms, respectively. The amplitude of the square wave signal is adjusted so that the optical power of the signals that pass through the AOS are modulated with a 100% modulation depth. Note that this change in the traffic pattern (17 channels are dropped/lost) may be associated with a 94.4% traffic drop or loss. Fig. 11(a) shows the response of one of the affected (lost) channels (channel 17), and Fig. 11(b) shows the response of channel 18 (the surviving channel) as it propagates through the network after each amplification stage. Fig. 11(a) also shows the duration of the disturbance in the traffic. The repetition rate is 100 Hz and the disturbance duration is 2 ms. The disturbance duration is defined as the time between the drop and the add of the 17 channels. Fig. 11 shows that the power excursions grow with each amplification stage. The power overshoots do not exceed 4.5 dB, while the power undershoots are less than 1.5 dB. Fig. 12(a) shows this increase in the magnitude of the overshoot and undershoot as a function of the number of the amplification stages. The magnitude of the power overshoot (undershoot) is defined as the difference of the maximum (minimum) power to the initial steady-state power value before the event. Fig. 12(b) shows the time it takes for the overshoot and undershoot to develop (reach their maximum value) after 17 channels have been suddenly dropped/lost. It can be seen that the maximum overshoot is reached after 0.05 ms while the maximum undershoot

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 6, JUNE 2002

is reached after 0.4 ms and they are independent of the number of amplification stages. However, for disturbances that last less than the settling time of the amplifier control mechanism, as the one shown in Fig. 13 (repetition rate 100 Hz, and disturbance duration of 0.2 ms), the response is different. We define the time it takes for the control loop and the amplifier to reach its steady state (within 5%) after the event (add–drop) as the settling time of the gain control mechanism. In this case, the gain-controlled amplifiers have not reached their steady state when the recovery of the 17 channels occurs. This causes larger power transients now as the control mechanism tries to converge. Nevertheless, the magnitude of these undershoots did not exceed 2.5 dB in our setup. We also examined the transient response for our system for other disturbance duration such as 0.5 ms and 1 ms. Fig. 14 shows the power excursions for disturbance duration of 0.2, 0.5, 1, and 2ms at the end of the path and after four amplification stages. For disturbance duration of 0.2 ms and 0.5 ms, the power undershoots are higher (e.g., 2.5 dB) than the undershoots for the other two cases (less than 1 dB). Note also that the maximum value of the overshoot occurs at the same time after the loss of the 17 channels and is independent on the disturbance duration and the repetition rate. On the other hand, the undershoot depends on the disturbance duration. If the disturbance duration is faster than the settling time of the amplifier, the gain control mechanism has not reached its steady state. Thus the maximum undershoot occurs after the recovery of the traffic while the control mechanism tries to converge. If the disturbance duration is slower than the settling time of the amplifier, the control mechanism has enough time to reach its steady state and the maximum undershoot occurs 0.04 ms after the loss of traffic. The above results are also shown in Fig. 15, where the magnitude of the maximum overshoot and undershoot and the time of occurrence of the transient as a function of the disturbance duration are plotted. Note that no significant change exists for disturbances that are much slower than the settling time of the amplifier, only in the case of 0.2 ms duration the power change and the occurrence of the maximum undershoot are significantly different from the other cases. Nevertheless, power transients observed for disturbance durations of larger than 1 ms do not significantly change the -performance of the network. This is shown in Fig. 16 where the -factor is plotted versus the disturbance duration for the two repetition rates. A -penalty of 0.5 dB is observed for a disturms and repetition rates of 100 Hz compared bance duration with the static case (no added/dropped traffic). However, when ms the is reduced the disturbance duration is reduced to (the penalty is increased). This degradation for shorter disturbance duration is due to the larger undershoots of the power excursion. When the traffic is recovered before the amplifier control settling time, larger undershoots are observed because they occur before the gain control amplifiers reach their steady state (typically within 1 ms). This is the reason why a higher -penalty is observed at duration values of 0.2 and 0.5 ms. As dB mentioned earlier, our system is OSNR limited (OSNR after 280 km); hence, large undershoots are more detrimental since they can cause the OSNR to drop and this reduces the factor. The factor was also measured for repetition rates of 500 Hz. A 1-dB -penalty was measured compared to the static case (no added/dropped traffic) for a disturbance duration

MADAMOPOULOS et al.: STUDY OF THE PERFORMANCE OF A METROPOLITAN AREA NETWORK

(a)

943

(b)

Fig. 12. (a) Magnitude of the maximum overshoot and undershoot as a function of the number of amplification stages. (b) Occurrence of the transient as a function of the number of amplification stages.

Fig. 14. Comparison of power overshoot and undershoots for channel 18 after four amplification stages. (Repetition rate: 100 Hz, Disturbance duration: 0.2, 0.5, 1, and 2 ms). Fig. 13. (a) The response time of the dropped/lost channel 17 with disturbance duration 0.2 ms. (b) The response of channel 18 showing the transients suffered by the sudden drop of 17 channels. (Repetition rate: 100 Hz.)

of 0.5 ms. In this case, the factor is lower than in the previous case because the optical power changes occur five times within the same time period when compared to the 100-Hz case. Hence, more errors are measured giving a lower estimated . V. CONCLUSION In conclusion, we demonstrated a transparent reconfigurable metropolitan area network testbed of total size 280 km and 45 Gb/s (18 channels at 2.5 Gb/s) capacity. The network consisted of three different node designs that allow for different degrees of add–drop functionality. Excellent transmission dB) was obtained for the longest path performance ( in the network for all channels tested. We showed that a significant change in the traffic load (e.g., 94% of the traffic is lost) degrades the -performance of the channels by 0.5 dB. This happens for repetition rates of 100 Hz and lasts for a

duration on the order of the settling time of the amplifier gain degradation increases at dB for control or larger. The disturbances occurring at a repetition rate of 500 Hz. This is due mainly to the fact that within the same measurement interval, as for the 100 Hz repetition rate, more disturbances occur and consequently more errors are observed. At these higher repetition rates and for disturbance durations smaller than the amplifier gain control settling time, the degradation in the -performance is rather significant. Note that the disturbances we chose to emulate in our network testbed occur in a periodic fashion. This is not something that is typical in optical networks, where it is expected that such disturbances will occur at a much slower rate and hence affecting the system performance less. From our results, we can also conclude that for disturbances leading to a smaller number of lost channels, smaller -degradation is expected. Our study shows that optical transparency is feasible in metro-area networks when proper optical components, fiber, and network elements are chosen. Future work includes studying larger networks with a larger number of amplifiers.

944

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 6, JUNE 2002

(a)

(b)

Fig. 15. (a) Magnitude of the maximum overshoot and undershoot as a function of the disturbance duration. (b) Time of occurrence of the transient as a function of the disturbance duration. No significant change exists for disturbances that are much slower than the settling time of the amplifier.

Q

Fig. 16. -factor as a function of the disturbance duration (or recovery time for the loss of signal) for repetition rates of 100 and 500 Hz.

ACKNOWLEDGMENT The authors would like to thank M. Lelic for useful discussions. REFERENCES [1] M. J. Riezenman, “Optical nets brace for even heavier traffic,” IEEE Spectr., pp. 44–46, Jan. 2001. [2] C. A. Brachelt, “Foreword: Is there an emerging consensus of WDM networking?,” J. Lightwave Technol., vol. 14, pp. 936–941, 1996. [3] A. M. Saleh and J. M. Simmons, “Architectural principles of optical regional and metropolitan access networks,” Lightwave Technol., vol. 17, pp. 2431–2448, 1999. [4] P. Leisching, H. Bock, A. Richter, C. Glingener, P. Pace, B. Keyworth, J. Philipson, M. Farries, D. Stoll, and G. Fischer, “All-optical-networking at 0.8 Tb/s using reconfigurable optical add/drop multiplexers,” IEEE Photon. Technol. Lett., vol. 12, pp. 918–920, 2000. [5] K.-D. Langer, J. Vathke, G. Walf, U. Broniecki, L. Dembeck, G. Eilenberger, R.-P. Baraun, A. Gladisch, S. Szuppa, H. Bock, P. Leisching, A. Richter, and W. Grupp, “Transparent interconnection of multi-vendor metropolitan sub-networks,” in Proc. ECOC’00, vol. 4, Munich, Germany, 2000, pp. 87–88. [6] T. Pfeier, H. Schmuck, B. Deppisch, M. Witte, and J. Kissing, “TDM/CDM/WDM approach for metro networks with 200 optical channels,” in Proc. ECOC’00, vol. 3, Munich, Germany, 2000, pp. 71–72.

[7] M. Morin, J. Chauvin, D. H. B. Hoa, L. Berthelon, and O. Rodal, “PanEuropean lightwave core and access network (ACTS pelican project),” in Proc. ECOC’00, vol. 4, Munich, Germany, pp. 85–86. [8] K. Y. Jarkko, S. Tammela, T. Niemi, and A. Tervonen, “Scalability of a metropolitan multi-fiber bidirectional WDM ring network,” in Proc. ECOC’00, vol. 3, Munich, Germany, pp. 81–82. [9] M. White, D. Wonglumsom, K. Shrikhande, S. M. Gemelos, M. S. Rogge, and L. G. Kazovsky, “The architecture of HORNET: A packet-over-WDM multiple-access optical metropolitan area ring network,” Computer Networks, vol. 32, pp. 587–598, Sept. 2000. [10] M. D. Vaugh and R. E. Wagner, “Metropolitan network traffic demand study,” in Proc. LEOS’00 Annu. Meet., vol. 1 MK4, Puerto Rico, pp. 102–103. [11] N. Madamopoulos, M. D. Vaugh, L. Nederlof, and R. E. Wagner, “Metro network architecture scenarios, equipment requirements and implications for carriers,” in Proc. OFC 2001, vol. WL2, Anaheim, CA. [12] I. Tomkos, R. M. Hesse, C. D. Friedman, N. A. Antoniades, N. Madamopoulos, B. S. Hallock, R. S. Vodhanel, and A. Boskovic, “Transport performance of a transparent WDM regional area ring network utilizing optimized components/fiber,” in Proc. OFC 2001, vol. PD35, Anaheim, CA. postdeadline paper. [13] N. Antoniades, A. Boskovic, J. Downie, N. Madamopoulos, D. Pastel, J. K. Rhee, I. Roudas, I. Tomkos, and M. Yadlowsky, “Engineering the performance of DWDM metro networks,” in Proc. NFOEC 2000, Denver, CO, pp. 204–211. Session B2: Metro DWDM Applications. [14] N. Madamopoulos, C. Friedman, I. Tomkos, and A. Boskovic, “Study of a transparent and reconfigurable metropolitan area network,” in Proc LEOS’01 Annu. Meet., vol. WD-2, La Jolla, CA, 2001, pp. 427–428. [15] P. J. Corvini and T. L. Koch, “Computer simulation of high bit rate optical fiber transmission using single frequency lasers,” J. Lightwave Technol., vol. LT-5, pp. 1591–1595, 1987. [16] J. C. Cartledge and G. S. Burley, “The effect of laser chirping on lightwave system performance,” J. Lightwave Technol., vol. 7, pp. 568–573, 1989. [17] I. Tomkos, R. Hesse, N. Antoniades, and A. Boskovic, “Impact of filter concatenation effects on the performance of metropolitan area optical networks utilizing directly modulated lasers,” in Proc. OFC 2001, vol. WBB4, Anaheim, CA, 2001. [18] C. C. Wang, I. Roudas, I. Tomkos, M. Sharma, and R. S. Vodhanel, “Negative dispersion fibers for uncompensated metropolitan networks,” in Proc. ECOC’00, vol. 1, Sept. 2000, pp. 97–98. [19] L. Noire, M. Vigoureux, and E. Dotaro, “Impact of intermediate traffic grouping networks,” in Proc. OFC 2001, vol. TuG 3, Anaheim, CA. [20] P. A. Humblet, “The direction of optical technology in the metro area,” in Proc. OFC 2001, vol. WBB1, Anaheim, CA, 2001. [21] R. Ranalli, B. A. Scott, and J. P. Kondis, “Liquid crystal-based wavelength selectable cross-connect,” in Proc. ECOC 1999, vol. 1, Nice, France, pp. 68–69. [22] Synchronous Optical Network (SONET): Physical Layer Specification. [23] Optical Interfaces for Equipment and Systems Relativity to the Synchronous Digital Hierarchy, July 1995.

MADAMOPOULOS et al.: STUDY OF THE PERFORMANCE OF A METROPOLITAN AREA NETWORK

[24] Y. Sun, A. K. Srivastava, J. L. Zyskind, J. W. Sulhoff, C. Wolf, and R. W. Tkach, “Fast power transients in WDM optical networks with cascaded EDFAs,” Electron. Lett., vol. 33, no. 4, pp. 313–314, 1997. [25] H. Richards, J. L. Jackel, and M. A. Ali, “A theoretical investigation of dynamic all-optical automatic gain control in multichannel EDFA’s and EDFA cascades,” IEEE J. Select. Topics Quantum Electron., vol. 3, pp. 1027–1036, July 1997. [26] G. Luo, J. L. Zyskind, J. A. Nagel, and M. A. Ali, “Experimental and theoretical analysis of relaxation-oscillations and spectral hole burning effects in all-optical gain-clamped EDFA’s for WDM networks,” J. Lightwave Technol., vol. 16, pp. 527–533, Apr. 1998. [27] M. I. Hayee and A. E. Willner, “Transmission penalties due to EDFA gain transients in add-drop multiplexed WDM networks,” IEEE Photon. Technol. Lett., vol. 11, pp. 889–891, July 1999. [28] K. Motoshima, D. N. Chen, L. M. Leba, M. M. Downs, T. Li, and E. Desurvire, “Dynamic compensation of transient gain saturation in erbium-doped fiber amplifiers by pump feedback control,” IEEE Photon. Technol. Lett., vol. 5, pp. 1423–1426, Dec. 1993. [29] A. K. Srivastava, Y. Sun, J. L. Zyskind, J. W. Sulhoff, C. Wolf, and R. W. Tkach, “Fast control in erbium-doped fiber amplifier,” in Tech. Dig. OAA’96, pp. 24–27. [30] M. Karasek and J. C. Van Der Plaats, “Analysis of dynamic pump-loss controlled gain-locking system for erbium-doped fiber amplifiers,” IEEE Photon. Technol. Lett., vol. 10, pp. 1171–1173, Oct. 1998. [31] M. Begin, J. Lauzon, Y. Rouleau, and Y. Mimeault, “Gain-locked dualstage EDFA for WDM systems,” in Tech. Dig. OAA’00, pp. 148–150. [32] H. Okamura, “Automatic optical loss compensation with erbium-doped fiber amplifier,” J. Lightwave Technol., vol. 10, 1992. [33] S. Y. Park, H. K. Kim, G. Y. Lyu, S. M. Kangnm, and S. M. Shin, “Dynamic gain and output control in gain-flattened erbium-doped fiber amplifier,” IEEE Photon. Technol. Lett., vol. 10, pp. 787–789, 1998. [34] N. Takahashi, T. Hirono, H. Akashi, S. Takahashi, and T. Sasaki, “An output power stabilized erbium doped fiber amplifier with automatic gain control,” IEEE J. Select. Topics Quantum Electron., vol. 3, pp. 1019–1026, 1997. [35] M. Lelic, A. Bauco, N. Menon, K. Wundke, and T. Zahnley, “Smart EDFA with embedded control,” in Proc. LEOS’01 Annu. Meet., La Jolla, CA, Paper WC-3, pp. 419–420.

Nicholas Madamopoulos (S’95–M’99) received the B.S. degree in physics (with honors) from the University of Patra, Greece, in 1993 and the M.S. and Ph.D. degrees in optical science and engineering at the School of Optics, Center for Research and Education in Optics and Lasers (CREOL), University of Central Florida, Orlando, in 1996 and 1998, respectively. After receiving the B.S. degree, he joined the Applied Optics and Optical Processing Laboratory of the Institute of Electronic Structures and Lasers, at the Foundation for Research and Technology-Hellas (FO.R.T.H.), Heraklion, Greece, as a Research Fellow. There, he conducted research on dynamic diffraction at the near infrared and excimer laser micromachining for computer generated holograms and microoptic diffractive elements. In August 1994, he received a Graduate Optics Fellowship at the Center for Research and Education in Optics and Lasers (CREOL). His Ph.D. specialization was in photonic information processing systems, where he introduced novel photonic delay lines for phased array antenna applications, as well as photonic processing modules for fiberoptic communications. He was a Graduate Research Fellow at the Photonic Information Processing Systems Laboratory at CREOL. In summer 1999, he joined the Network Architecture and Equipment Department, Corning, Inc., in Corning, NY, where he was concentrated on modeling the optical performance of the physical layer of optical network architectures. In fall 2000, he joined the Corning Photonics Research and Test Center facility, Somerset, NJ, where he is mostly involved with experiments. His research interests are optical network architectures and transport layer performance of WDM networks. Dr. Madamopoulos is a Member of IEEE-Lasers and Electro-Optics Society and the Optical Society of America (OSA). He is a Reviewer for IEEE and OSA publications. He received a New Focus Student Essay Prize in 1996, the SPIE Educational Scholarship in Optical Engineering in 1997, a Graduate Merit Fellowship Award in 1998 and a New Focus/OSA Student Award in 1998. He was one of the founding members of the first IEEE-LEOS Student chapter and he served as treasurer and president for several years.

945

D. Clint Friedman was born in New Jersey in 1977. He received the B.S. degree in physics from Rutgers University, New Brunswick, NJ, in 2000. He is currently pursuing the M.Sc. degree in engineering physics (optics) at Stevens Institute of Technology, Hoboken, NJ. After receiving the B.S. degree, he joined the Photonic Research and Test Center, Corning, Inc., Somerset, NJ. He is involved in performance characterization of optical network elements and components supporting various Corning businesses.

Ioannis Tomkos (S’98–A’99–M’02) received the B.S. degree in physics from the University of Patras, Greece, and the M.Sc. degree in telecommunications engineering and Ph.D. degree in optical telecommunications from the University of Athens, Greece, in 1994, 1996, and 1999, respectively. In 1996, he joined the Optical Communications Group of University of Athens, Greece, where he participated in several national and European research projects (e.g., ACTS, COST) as a Research Fellow. His work there was related with WDM technologies for all-optical networks (including characterization of optical components, all-optical signal processing, design of novel optical cross-connects, short-pulse generation, and dispersion compensation) and with digital transmission systems for access networks (e.g., ADSL, HFC technologies). His Ph.D. work focused on theoretical and experimental studies of novel wavelength conversion technologies based on nonlinear effects in semiconductor optical amplifiers/lasers and fibers. During his Ph.D. studies he was a Visiting Researcher in several leading research centers across Europe. In January 2000, he joined the Photonics Research and Test Center of Corning, Inc. as a Senior Research Scientist. His research focused in theoretical and experimental studies of transport phenomena related with WDM optical networks. He studied extensively the performance and design issues of metropolitan/regional area optical networks. He is currently involved in studies related with transport performance and technology trends in ultralong-haul optical networks. He is author or coauthor of about 60 contributed and invited papers, published in international journals and conference proceedings and he has several patent applications pending. Dr. Tomkos is a Member of IEEE-Lasers and Electro-Optics Society. In 1991, 1992, and 1993, he received an Educational Scholarship by the Hellenic Institute of National Scholarships. In 1998, he received a Student Grant and also the “Best Student Paper” Award from IEEE LEOS. In 1999, he received an Educational Scholarship from SPIE. In 2000, he was awarded by Corning Inc. for his contributions in the development of a new optical fiber. In 2002, he received the 2001 Corning research Outstanding Publication Award. He is a Reviewer for the JOURNAL OF LIGHTWAVE TECHNOLOGY and the IEEE PHOTONICS TECHNOLOGY LETTERS.

Aleksandra Boskovic received the B.S. and M.S. degrees in physics from the Pontifical Catholic University of Rio de Janeiro, Brazil, in 1988 and 1991, respectively, and the Ph.D. degree in physics from the University of London, Imperial College, U.K., in 1996. She joined the Science and Technology Division, Corning Incorporated, Corning, NY, as a Senior Scientist in 1997, and since then, has worked on optical fiber and components transmission system impairments and optical networking in general. She is currently a Manager at Corning Incorporated, leading a group in Optical Network Performance and Value Proposition. Dr. Boskovic is a Member of the Optical Society of America (OSA) and the Optical Fiber Communication Conference (OFC’03) Committee on “Networks: Switching, Access, and Routing.”