We experimentally investigate the performance of burst-mode EDFA in an optical packet and circuit integrated system. In such networks, packets and light paths can be dynamically assigned to the same fibers, resulting in gain transients in EDFAs throughout the network that can limit network performance. Here, we compare the performance of a ‘burst-mode’ EDFA (BM-EDFA), employing transient suppression techniques and optical feedback, with conventional EDFAs, and those using automatic gain control and previous BM-EDFA implementations. We first measure gain transients and other impairments in a simplified set-up before making frame error-rate measurements in a network demonstration.
©2013 Optical Society of America
The new-generation of optical networks are expected to provide a wide variety of data communication services, each with different network requirements. For example, applications such as web-browsing e-mail or distribution of data from sensor networks require only a modest bandwidth that may be met with best-effort services. However, demanding applications such as digital cinema or telemedicine will require communication services with both huge bandwidth and guaranteed end-to-end quality of service (QoS).
We have been developing an optical packet and circuit (OPCI) node system  that integrates optical packet switching (OPS) and optical circuit switching (OCS) into one optical network infrastructure. OPCI is designed to simultaneously support multiple communication services with different network requirements . In OPCI nodes, OPS technology is used to deliver a best-effort service with OCS used to provide full light-path provisioning, providing both large bandwidth and the guaranteed QoS. Furthermore, using dense-wavelength division multiplexing (DWDM) technologies, OPCI can perform the dynamic re-allocation of wavelength resources between OPS and OCS to adapt to changing traffic demands. In the OPCI node, optical amplifiers are employed not only for transmission but to compensate for losses of optical signal processing within the nodes with common paths usually used for both path and packet signals. Hence, it is imperative that EDFAs be able to simultaneously amplify signals with varying input power without impairment from input-power fluctuations (IPFs) that can result in gain transients [3,4], that may cause serious signal degradation and limit network performance. Furthermore, OPCI networks may be dynamically reconfigured according to network load so EDFAs in such networks must also be able to cope with in a range of traffic loads, some of which also place a requirement for high gain on amplifiers.
A common approach for reducing gain transients and their impact is to employ electrical automatic gain control (AGC) [5–8] or optical gain clamping by optical feedback (OFB) [9,10]. However, AGC-EDFAs may be limited by the speed of the feedback circuit, which may be of the order of microseconds, making it insufficient to control the packets around 100ns duration or sharp power fluctuations. Furthermore in an OPCI system, changing the amplifier gain to compensate for changes of packet signals will change the gain seen by OCS channels leading to undesirable transients on those signals. Optical feedback can clamp the gain of optical signals with varying input power, but trade-off the range of tolerable IPFs with the total gain of the amplifier. To address these issues, we previously proposed transient suppressed EDFA (TS-EDFA) [11–13], designed with large intrinsic saturation power which was previously demonstrated to provide high gain even with OFB [14–16].
In this paper, we investigate the performance of EDFAs in OPCI networks, where long and short optical packet signals and light-path signals are transmitted along the same fiber simultaneously. We first use a simplified network set-up to compare the magnitude of gain transients from several EDFA variants, including conventional EDFAs, an EDFA with AGC and TS-EDFA, with our latest burst-mode EDFA (BM-EDFA) that employs a TS-EDFA specifically designed with large intrinsic saturation power (PIS) for use with OFB. We see that the BM-EDFA has reduced susceptibility to gain transients and relaxation oscillations, in the presence of large IPFs compared to comparable EDFA technology. We then investigate the performance of each EDFA type in an OPCI network test-bed by performing data error rate measurements and again observe best performance with the BM-EDFA. These results show that the BM-EDFA is able to mitigate the impact of gain transients in a realistic OPCI scenario and can help enable successful implementation of such dynamic network architectures.
2. Overview of Theory
All the quantities in the gains are constant. And the gain at all signal channels are controlled through β. an increase of the total signal input power is automatically compensated by a decrease of the power in the feedback loop. It is evident that this gain control mechanism works only for conditions permitting lasing in the feedback loop. As a consequence, for a given value of β, there is a maximum signal input power above which gain stabilization can no longer be achieved.
3. Characterization of Burst-Mode EDFA
Before evaluation of the Burst-mode EDFA is the OPCI network, we first characterized the BM-EDFA and 4 other EDFAs in a simplified packet generation set-up to measure the magnitude of gain transient and relaxation oscillations as shown in Fig. 1. The device under test (DUT) receives the traffic as 10-wavelength channel optical packet signal and a continuous signal on a single wavelength channel. The optical packets were generated from 10 continuous wave (CW) lasers, multiplexed in an arrayed waveguide grating (AWG) and modulated with 10 Gbps non-return to zero (NRZ) pseudo random bit sequence (PRBS) of length 29-1 in a LN modulator. A synchronized packet signal was then used to superimpose the packet envelopes using an acousto-optic modulator (AOM). The feedback loop consists of a variable optical attenuator (VOA3), used to control the loop attenuation, 1/β and a 0.8nm optical bandpass filter to select the feedback wavelength. The feedback wavelength was set at 1530nm to take advantage of the natural EDFA gain peak allows a reduction of the loop coupling losses maximizing output power for signal channels, as described in . Furthermore, to maintain equal gain between channels, wavelength around the 1530nm gain peak is currently used in the OPCI network. Before the receiver, a second 0.8nm filter was used to select the channel for direct measurements of gain transients and Q factor variations which was performed in a digital sampling oscilloscope (DSO).
Using this set-up, we compared the characteristics of five EDFAs variants. DUT1 and 2 are conventional EDFAs from different manufacturers, DUT3 is a TS-EDFA, DUT4 is an electrical AGC-EDFA, and DUT5 is the BM-EDFA consisting of a TS-EDFA with OFB. The small signal gains of each are shown in Fig. 2. Also shown is the gain as a function of loop attenuation for DUT1-3 with OFB. We observe that the small signal gain is decreased by about 10dB compared to the case without OFB for a loop attenuation value of 11dB. OFB does slightly increase the noise figure in a conventional EDFA , but has not yet been studied in TS-EDFA.
Firstly, we observed gain transients and eye diagrams by setting VOA2 to provide the same POUT(λj) for each of the DUTs. The magnitude of the gain transient is a function of POUT as shown in Eq. (1). Therefore the comparison of those of each EDFA is possible under the same POUT condition. We then measured the Q-factor and gain transient of the optical packet and adjusted the timing signals to vary the packet rate between 0.1 and 50% and the packet length between 137nm and 16.4μs. Figure 3 shows the Q-Factor value measured in each case, while Fig. 4 shows the amplitude variation and the eye diagrams of each DUTs when optical packet length of 1.64μs and 1% packet rate signal are used. From Fig. 4, it is evident that the presence of gain transients broadens the distribution of the eye diagrams and hence reduce the Q-factor of the received signals.
The conventional EDFAs, DUT1 and 2, show a smaller measured Q-factor than the other DUTs and gain transients can be observed in the packet waveforms. With OFB employed, the Q-Factor of packets traversing these amplifiers improves, but at the cost of the amplifier gain. For DUT4, it appears that the electrical AGC works well and the highest Q-Factor is measured for micro second packet lengths. However when the packet length reduces to 137ns (64byte 10GbEther frame time) the Q-factor of packets transmitted through DUT4 is decreased because of insufficient speed of the electrical AGC control. DUT5, the burst mode EDFA, has both negligible gain transient and large stable Q-factors overall. However, the use of optical feedback also introduced additional amplitude fluctuations due to relaxation oscillations associated with the laser cavity .
Figure 5 shows relaxation oscillation of the surviving channel when the signal switch frequency of the dropped channel is 1 kHz. The gain fluctuation caused by relaxation oscillation of DUT1 with OFB was observed to be about 1dB even in the case of 10dB loop attenuation (red line). Alternatively, DUT5, showed fluctuations caused by relaxation oscillations to be under 0.3dB.
Figure 6 shows the measured gain transient of the surviving continuous channel. For DUT 1 and 2, the magnitude of the gain transient on the surviving channel is roughly equal to the power of the dropped channel. There is little oscillation on the surviving channel for DUT 3, 4 and 5 when either long or short packet signals are used. Although the magnitude of relaxation oscillations is small compared to the dominant signal degradation mechanism of steady state power fluctuations, these results show that the BM-EDFA design is also able to suppress relaxation oscillations when OFB is used.
4. Experiments with the OPCI network
Next, we investigated the transmission characteristics of each DUT on the optical packet and circuit integrated network. Figure 7 shows the experimental setup for transmission performance measurements. The total size of the10GbE frames of the interface with the client-side network in the optical packet and circuit integrated network are 64, 1518, 9000 bytes. The optical packet sub-system transmitted 10-wavelength, 100Gbps optical packets of 157ns length, carrying 1518 bytes as commonly used in electrical packet switching. In the OCS sub-system, the optical signals are monitored and controlled to be a constant averaged power. However, for optical packet transmission in OPCI nodes, the signal average power depends on the packet rate of burst signals. Hence, the measurements reported here use a constant peak power for packet channels.
Node1 and Node2 are OPCI nodes. We transmitted 10GbE frames by 14 wavelength optical packets and continuous light-paths on 10 wavelengths. The signals, transmitted from Node1 to Node2 are split in a coupler before the DUT to enable monitoring of the DUT input signal, enabling selection of the number of optical circuit wavelength channels the DUT receives by adjustment of the optical filter. The output coupler provides the output signals of the DUT to both the optical packet receiver (100G-OP transponder) and the optical circuit receiver (10G OTN) of node 2. The OPS signals are combined with an amplified spontaneous emission (ASE) noise source to enable control of the OSNR. Similarly, the signal sent to the 10GOTN and 100G-OP were also noise loaded to allow BER measurements as a function of the OSNR.
The length of an optical packet was variable, corresponding to the length of the 10GbE frames . Table 1 shows the setting of packet frames of the router tester. We fixed the 10GbE frame length to be 1518 bytes, and the optical packet rates used were set at 9.7% (High density) and 0.013% (low density) by adjusting the inter frame gap of 10GbE frames. Figure 8 shows the wave forms of the transferred 100G-OP signals and the spectrum of received signals of DUTs.
Figure 9 shows the measurement results of output wavelength channel spectrum of each DUTs when the received wavelength channel of packet signal was at the high condition and a variable number of OCS wavelengths, from 14 (black line), 10 (green), 4(red line), to 1 (light blue). Figure 8 shows that as the number of OCS channel is reduced, the power of packet signals increases by over 3dB. Such power fluctuation force some signals outside of the dynamic range of the 100G-OP and 10GOTN receivers and hence degrade the BER of transmitted signals.
Under the same conditions, Fig. 10 shows that the opposite effect for DUT 4, where the AGC actually increases the power of the packet channel by up to 1.5dB as the number continuous channels is increased from 1 to 14. However, the combination of TS-EDFA and OFB is able to tolerate the large change of input power without any noticeable impact on the packet channel power.
Finally, we measured the frame error rate (FER) in an OPCI network demonstration to quantify the impact of gain transients on network performance. Figure 11 shows the results of measured FER of the optical packet signals as a function of the OSNR. The legend of Fig. 11 shows number of OCS channels. It was described “DUT1 path1” when DUT1 received 1 OCS wavelength and 10 packet signals. We observe that the number of OCS wavelengths present can have an impact on the FER, particularly for DUT1 to 3 and for the low density packet signal. However, the FER of DUT5 is independent of the number of OCS channels since the power of the packet channel is unaffected by the adding and dropping of OCS wavelengths. The FER performance of DUT4, has similar characteristics with DUT5 for the high-density packet scenario, but maybe degraded by up to 2dB by OCS channel add/drop in the low traffic density case. In some cases for few OCS channels and DUT 1 to 3, the magnitude of gain transients made FER measurements impossible and hence these are crossed out in the legend of Fig. 11.
Figure 12 shows the measured FER of survived path signal after nine out of ten channels are dynamically added and dropped. Since the received signal power was adjusted to a reference value before the steady state measurement, it was not so easy to detect the impact of OCS channel add/drop on the FER. However, Fig. 12 shows that when the number of wavelength channel of light-path dynamically changed, the FER is degraded by the received power change of each EDFAs with the exception of DUT5. These results show that the impact of gain transients may not be strong in conditions of high traffic density when amplifiers are typically saturated. However, for a fully dynamic network, it must be able to deal with all required traffic demands and these results show that it is important to characterize such networks to determine under which conditions these gain transients are problematic. Such investigation allows amplifier characteristics, such as optical feedback cavity and saturation power, to be designed accordingly and to provide sensible planning guidelines . Overall, we show that the BM-EDFA is able to mitigate the impact of gain transients in a wide range of scenarios demonstrating that the burst mode EDFA is a key component for future OPCI networks.
We developed a burst mode EDFA employing a transient suppression technique with optical feedback. The EDFA was shown to minimize gain transients, relaxation oscillations, and data error rate in transmission in the test-bed OPCI network. By comparison, conventional EDFA (DUT1, 2) were observed to exhibit strong gain transients and relaxation oscillations, which could be reduced by employing optical feedback but at the expense of a large reduction in available signal gain. A previous version of TS-EDFA (DUT3) was observed to cause degradation of the FER when the number of wavelength channel of light-path changed. And the EDFA with electrical AGC (DUT4) causes a gain fluctuation when the packet signals density is low. The BM-EDFA was obtained stable gain characteristics in a range of measurements demonstrating its potential as a key component of OPCI networks.
We would like to thank M. Kurihara, T. Hashimoto, H. Sumimoto, and T. Makino of the National Institute of Information and Communications Technology for their support in the experiments.
References and links
1. N. Wada, “Photonic technologies for new generation network,” Communications and Photonics Conference and Exhibition (ACP) 2010 Asia, pp.463–464 (Shanghai, China, December, 2010).
2. H. Harai, “Optical packet and circuit integrated networks,” in 17th International Conference on Optical Network Design and Modeling (Brest, France, April, 2013), pp.76–81.
4. B. Neto, A. Klingler, C. Reis, J. P. Girao, A. Teixeira, and P. S. Andre, “EDFA transient assessment for bursty traffic,” in 3rd ICTON Mediterranean Winter Conference MW2009, (Angers, France, December, 2009), no.FrP.6.
5. C. Tian and S. Kinoshita, “Analysis and control of transient dynamics of EDFA pumped by 1480- and 980-nm lasers,” J. Lightwave Technol. 21(8), 1728–1734 (2003). [CrossRef]
6. H. Nagaeda, Y. Horiuchi, Y. Tanaka, N. Shiga, Y. Fukada, K. Suzuki, N. Yoshimoto, and M. Tsubokawa, “Burst-mode optical amplifier for signals with various burst lengths from 500 ns to 50 us,” in Proceedings of ECOC (Brussels, Belgium, September, 2008), no.P.1.13.
7. N. Sato, K. Ota, N. Mishima, Y. Oikawa, and N. Shiga, “Less than 0.19-dB transient gain excursion AGC-EDFA with digital control for 20-channel add/drop equivalent operation,” in Proceedings of OFC/NFOEC (Los Angeles, U.S.A., March, 2011), no.OMH3.
8. K. Ishii, J. Kurumida, K. Tanizawa, S. Namiki, N. Sato, K. Ota, Y. Oikawa, and N. Shiga, “Suppression of transients in an EDFA chain using feed-forward pump control and a high-speed VOA,” in Proceedings of OFC/NFOEC2012 (Los Angeles, U.S.A., March, 2012), no.OW4D.
9. M. Zirngibl, “Gain control in erbium-doped fibre amplifiers by an all-optical feedback loop,” Electron. Lett. 27(7), 560–561 (1991). [CrossRef]
10. M. Zannin and K. Ennser, “Impact of burst size and inter-arrival time in all-optical gain clamping amplification for optical burst switched networks,” J. Lightwave Technol. 31(6), 855–859 (2013). [CrossRef]
11. Y. Awaji, H. Furukawa, and N. Wada, “Impairment in amplification of optical packets regarding the gain transient and nonlinear effect depending on peak power of NRZ payload,” in Proceedings of OECC2007 (Shanghai, China, October, 2007), no.PDP2–4.
12. Y. Awaji, H. Furukawa, N. Wada, E. Kong, P. Chan, and R. Man, “Guidelines for amplification of optical packets in WDM environment regarding impact of transient response of erbium-doped fiber amplifier,” Comput. Netw. 52(10), 2087–2093 (2008). [CrossRef]
13. B. J. Puttnam, Y. Awaji, and N. Wada, “Investigating the limits of optical packet transmission through cascade transient-suppressed EDFAs without regeneration or active gain control,” in Proceedings of OFC/NFOEC2010 (San Diego, U.S.A., March, 2010), no.OThI6.
14. B. J. Puttnam, Y. Awaji, and N. Wada, “Investigating the performance of a transient-suppressed EDFA in optical packet and burst-switched networks,” IEICE Trans. Commun. E94-B(7), 1853–1859 (2011). [CrossRef]
15. B. J. Puttnam, Y. Awaji, and N. Wada, “Performance of an advanced transient suppressed EDFA in diverse dynamic optical network scenarios,” in Proceedings of ECOC2009 (Vienna, Austria, September, 2009), no.Tu 5.3.2.
16. B. J. Puttnam, B. C. Thomsen, A. Lopez, and P. Bayvel, “Experimental investigation of optically gain-clamped EDFAs in dynamic optical-burst-switched network,” J. Opt. Netw. 7(2), 151–159 (2008). [CrossRef]
17. D. 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. Sel. Top. Quantum Electron. 3(4), 1027–1036 (1997). [CrossRef]
18. G. Luo, J. Zyskind, Y. Sun, A. K. Srivastava, J. W. Sulhoff, C. Wolf, and M. A. Ali, “Performance degradation of all-optical gain-clamped EDFA’s due to relaxation-oscillations and spectral-hole burning in amplified WDM networks,” IEEE Photon. Technol. Lett. 9(10), 1346–1348 (1997). [CrossRef]
19. H. Furukawa, H. Harai, T. Miyazawa, S. Shinada, W. Kawasaki, and N. Wada, “Development of optical packet and circuit integrated ring network testbed,” Opt. Express 19(26), B242–B250 (2011). [CrossRef] [PubMed]
20. R. S. Luís, H. Furukawa, N. Wada, and S. Shinada, “Optical link planning guidelines for a hybrid packet and circuit integrated network,” in Proceedings of OFC/NFOEC2013 (Los Angeles, U.S.A., March, 2013), no.JTh2A.