TITLE: 100G COHERENT SYSTEM INTEROPERABILITY

TITLE: 100G COHERENT SYSTEM INTEROPERABILITY Pavel Škoda CESNET z.s.p.o. Zikova 4, Praha, Czech Republic Czech Technical University, Technická, Praha, Czech Republic e-mail: pavel.skoda@cesnet.cz Jan Radil CESNET z.s.p.o. Zikova 4, Praha, Czech Republic e-mail: jan.radil@cesnet.cz Josef Vojtěch CESNET z.s.p.o. Zikova 4, Praha, Czech Republic e-mail: josef.vojtech@cesnet.cz Miloslav Hůla CESNET z.s.p.o. Zikova 4, Praha, Czech Republic e-mail: m.hula@cesnet.cz Paper type Technical paper Abstract This paper presents results of laboratory and field testing of coherent 100G system with DP-QPSK modulation. The system was found quite resilient during laboratory tests that included power budget, nonlinear threshold, spectrum filtration, constellation diagram, interoperability with 10G lambdas and dispersion compensation type impact. Field tests addressed transmission of 100G signal as an Alien Wavelength through multivendor network, influence of photonic service parallel to 100G signal and performance of 100G system over single fiber bidirectional transmission lines. 100G system has been found extremely resilient to most classical impairments thanks to advances error coding and compatible with standard 10G NRZ lambdas and any type of dispersion compensation. The system was also working over single fiber bidirectional lines and in parallel with Photonic Service of time transfer. Keywords 100 Gbps, DP-QPSK, Field Tests, Dispersion Compensation, Photonic Service Introduction Coherent transmission formats pave the new era of optical communications and allow following Shannon prediction of system capacity increase [1]. Although many experts were doubtful about universal modulation for 40G technology, in case of the successive 100G technology they seem to unite for one candidate [2]. The Dual Polarization Quadrature Phase Shift Keying (DP-QPSK) modulation was also employed by Alcatel-Lucent (ALU) into their 100G coherent platform. New devices for optical coherent communication utilize powerful forward error coding mechanisms that considerably enhance transmission performance while eliminating the most of linear channel effects. One of the most important features questioned was interoperability with old 10G Non-Return-to-Zero (NRZ) modulation formats that require dispersion compensation through dispersion compensation fibers or Fiber Bragg Grating (FBG) modules. Also interoperability of coherent technology with new network applications or network equipment from different vendors is of wide interest. First chapter briefly introduces the most important equipment and facilities used during tests. Second chapter describes laboratory tests that identified basic parameters of the system. Third chapter presents tests at experimental laboratory fibre line addressing systems interference and dispersion compensation issues. Fourth chapter concerns field tests in multivendor network of CESNET2, special transmission over single fibre lines and parallel to a Photonic Service. 1 Used equipment and facilities

ALU 1830PSS Photonic Service Switch Platform The Alcatel-Lucent 1830 Photonic Service Switch (PSS) is the metro/regional Wavelength Division Multiplexing (WDM) platform purpose-built for flexible and automated WDM networking. New 100Gb/s platform signals rely on complex Dual Polarization Quadrature Phase Shift Keying (DP-QPSK) coherent modulation that fits into the 50 GHz ITU frequency grid. Signal resiliency to transmission impairments is heavily supported by Advanced Forward Error Coding (AFEC) mechanism that makes signals resistant to most negative linear effects as chromatic dispersion or polarization mode dispersion. EXFO PSO-200 Optical Modulation Analyzer The EXFO PSO-200 Optical Modulation Analyzer is designed for testing of transceivers of large variety of coherent modulation formats. The device coherently detects modulation signals and displays constellation and eye diagrams of received signals as well as their most important parameters. PSO-200 does not support forward error codes and therefore displays modulation symbols rather than transmitted data. APEX-T AP2443B Optical Complex Spectrum Analyzer The APEX-T AP2443B Optical Complex Spectrum Analyzer achieves unmatched resolution of 0.04 pm (5 MHz) with a close-in dynamic range of 60 db. Such excellent device parameters are achieved thanks to new interferometric principle used in this next generation device. Cisco ONS 15454 Multiservice Transport The Cisco ONS 15454 Multiservice Transport Platform (MSTP) is the most deployed metropolitan-area (metro) and regional dense wavelength division multiplexing (DWDM) solution in the world featuring twothrough eight-degree reconfigurable optical add/drop multiplexer (ROADM) technology that enables wavelength provisioning across entire networks and eliminates the need for optical-to-electrical-tooptical (OEO) transponder conversions. The ONS 15454 MSTP interconnects with Layer-2, Layer-3 and storage area network (SAN) devices at rates up to 40 Gbps. It delivers any service type to any network location and supports all DWDM topologies. CzechLight OpenDWDM The CzechLight OpenDWDM is a complete family of network devices for optical core networks designed to increase adaptability and flexibility of a core network to suit demmanding requests of Research and Education (R&E) community. CESNET uses OpenDWDM at about 40% of backbone network. CESNET2 network The CESNET2 network is the production part of the CESNET optical core network utilizing systems and devices from Cisco Systems. Cisco

DWDM equipment covers about 60% of CESNET optical core network with the biggest cities. Both Cisco DWDM and OpenDWDM are runing without any errors in a multivendor environment. The network performace is monitored at http://www.ces.net/netreport/. Optical laboratories of CESNET The optical laboratories of CESNET play the crucial role in applied research of CESNET. CESNET put considerable effort into research activities to improve flexibility of optical core network and so answer requests of R&E community. Up to date is CESNET performing module-to-system research in cooperation with Czech universities and research institutions. 2 Initial Laboratory Tests 2.1 Power budget The power budget of 100G DP-QPSK system from Alcatel-Lucent was tested in order to obtain maximum attenuation that can be overcome without optical amplification. A pseudorandom sequence from Bit Error Rate Tester (BERT) was multiplexed in ALU system onto 100G coherent signal. 100G signal passed only variable attenuator before demultiplexing and retrieving pseudorandom sequence for BERT, see Figure 1. Figure 1: Power budget test setup Thanks to Advanced Forward Error Coding (AFEC) embedded in ALU system, the system was able to overcome attenuation of almost 3 orders of magnitude (28.2 db) without any amplification. 2.2 Filtration test Filtration of optical spectra happens at many optical elements throughout every network. Also a slight misalignment of central wavelengths of optical elements is a common problem and therefore the resilience of system to the filtration effect is an important parameter. 100G ALU system was fed by pseudorandom sequence from BERT. 100G signal was then double filtered by two tunable filters ( A and B ), before it was sent back to 100G ALU system. The test setup is at Figure 2. Optical spectra were analyzed by high resolution spectrum analyzer from APEX Technologies. Figure 2: Setup of filtration test 100G signal was filtered by 100 GHz, 50 GHz and 30 GHz filters that suppressed mainly spectrum sidebands as can be seen in the left column of Figure 3. It is clear that information is mainly carried in central band and the signal fits into 30 GHz channel. Then 30 GHz filtering was done asymmetrically by detuning filter 40 GHz to the left and right. Detuning of 40 GHz was the maximum for a single filter before BERT experienced errors. At last the spectrum of double filtered signal at its central wavelength is displayed at the bottom of right column of

Figure 3. There was no impact to transmission BER for our filtering scheme as long as filter detuning did not increase beyond 40 GHz. The AFEC performance was not monitored due to time constrains. Figure 3: Spectra of 100G signal with applied filtering 2.3 Nonlinear effect threshold Once power in optical fiber increase over certain limit, nonlinear effects come into play. The most significant effect for 100G modulated signal in our setup is a self-phase modulation (SPM) that broadens signal spectrum while adding spurious phase modulation to signal. According to Figure 4, 100G signal from ALU system was amplified by booster amplifier and sent through 100km of G.652 fiber. Half of the signal was sent back to ALU system and the rest to spectrum analysis by APEX Technologies OSA. Figure 4: Setup for nonlinear threshold test The output power of optical amplifier was increased from 10dBm to 27dBm. BERT didn t report any errors until output power of 22dBm. Optical spectra for several amplifier output powers are at Figure 5. It can be seen that for powers over 20dBm SPM induces considerable broadening of main peak and vanishing of sidebands. Therefore amplifiers with high output power should be used with careful planning.

Figure 5: Nonlinear spectrum broadening as a function of booster output power 3 Experimental Transmission Line Tests 3.1 Transmission test Experimental transmission line in optical laboratory had length of 600km composing from both G.652 and G.655 fiber spans. Parameters of fiber spans can be found in Table 1. The test was set up using Dispersion Compensating Fibers (DCF) and long spans to explore performance of ALU system at older transmission links with DCFs. The initial setup with DCF as dispersion compensation was then replaced with Dispersion Compensating Modules (DCM) that are utilizing fiber Bragg Gratings (FBG) to prove that 100G technology can work together with standard systems. Two 100G signals from ALU systems were combined with 12 standard 10G NRZ channels in multiplexer and sent to the first amplification stage. The spectrum after multiplexer is at Figure 7. Double stage amplifiers with gain flattening feature were used in this experiment. All amplifiers except the first had dispersion compensating element in between their stages. The whole experimental transmission line can be seen at Figure 6. ID Type Length [km] Att [db] CD [ps/nm] A1 G.652c 100,7 20,1 1611,2 A2 G.652c 103,7 20,7 1659,0 A3 G.652c 101,1 20,2 1618,0 B1 G.655 95,4 19,1 19,4 B2 G.655 100,3 20,1 401,2 B3 G.655 100,6 20,1 402,4 Table 1: Used fiber spans in laboratory experiment

Figure 6: Experimental setup of 600km long transmission Figure 7: Spectra of 2x100G and 12x10G after initial multiplexer Both 10G and 100G systems worked without errors with proper dispersion compensation. In case of DCFs were systems working up to 844 ps of uncompensated dispersion without errors. Once some of DCF were changed for DCM, the error free operation dropped below 800 ps of uncompensated dispersion. For similar 814 ps BERT experienced some errors. So the conclusion drawn from this test is that FBGs have advantage in their length when compared to DCFs, but the dispersion must be compensated more accurately. This effect is probably caused by complicated dispersion compensating profile of FBGs. 3.2 Interference test Mutual influence of nearby optical channels is undesirable in DWDM systems. Therefore inter-channel interference test is of high importance for optical network planning. The experimental fiber line from Figure 6 was lit by 12x10G NRZ optical systems at 100 GHz grid and two 100G channels at 50 GHz grid in between the 10G channels. The performance of one 100G channel was observed by monitoring Advanced Forward Error Correction (AFEC) in 15 minutes intervals. AFEC experienced 1144 millions of corrected frames for adjacent 12x10G channels in system as can be seen on lower (red) figure of Figure 8. The same measurement has been done with three adjacent 10G channels removed to lower the interference, see upper (blue) figure of Figure 8, but suddenly the AFEC corrections rose up to 2282 millions of corrected frames. Although this effect does not follow common sense and is still under study, it is believed that this drop in system performance can be attributed to change of transmission power balance in combination with automatic power control of amplifiers along the experimental transmission line.

Figure 8: Spectra of interference test without adjacent 10G channels (blue), with adjacent 10G channels (red) 3.3 Constellation diagram test An eye diagram represents an effective way of displaying simple modulation schemes. A skilled engineer can easily judge signal quality and its parameters. More complex modulations like DP-QPSK are rather represented by constellation diagrams that show detected symbols in the complex plane. Dual Polarization Quadrature Phase Shift Keying (DP-QPSK) is a coherent modulation exploiting both phase and polarization to represent modulation symbols. Transceivers work at symbol rate of about 25 GBaud/s that translate to about 50 Gbit/s for four states QPSK in one polarization. Second polarization carries the same amount of information resulting in total throughput of about 100 Gbit/s. The optical coherent modulation analyzer PSO-200 from EXFO can detect and display various advanced optical modulation schemes. Analyzer is dedicated for coherent transceiver testing in production companies and therefore unable to correct transmission impairments. ALU transceiver was analyzed with PSO-200 and the clear constellation diagram of 100G DP-QPSK is at Figure 9. The figure displays constellation diagrams for both orthogonal polarizations (X polarization in up and Y polarization is below) and related eye diagrams of demodulated bit streams at detector. As PSO-200 does not support forward error codes the displayed values represents modulation symbols rather than actual data. Figure 10 shows constellation diagram for ALU transceiver with 10 km of G.652 fiber with precise DCF to compensate chromatic dispersion. The fiber length was prolonged to 100 km of G.652 and tunable dispersion compensation module based on FBG was used to optimize performance of PSO-200. A partial signal distortion can be seen at Figure 11 due to accumulated ASE from amplifiers in setup and some PMD. Figure 12 displays performance of PSO-200 in presence of 170 ps of uncompensated chromatic dispersion. Real long haul fiber lines usually experience more noise from multiple amplifier stages and larger PMD then fiber spools that prevented us from constellation analysis of real lines. It has to be highlighted that PSO-200 is not designed for system testing.

Figure 9: Constellation diagram directly after transceiver 100G DP-QPSK Figure 10: Constellation diagram of 100G DP-QPSK modulation after 10km of fiber with precise dispersion compensation

Figure 11: Constellation diagram of 100G DP-QPSK modulation after 100km of fiber with precise dispersion compensation Figure 12: Constellation diagram of 100G DP-QPSK modulation after 100km of fiber with 170ps uncompensated dispersion

4 Field tests in CESNET 2 live network Extensive laboratory tests were followed by several test scenarios in live network of CESNET 2 to verify interoperability of coherent 100G DP-QPSK system with standard 10G NRZ systems and special Photonic Services. Figure 13 shows CESNET2 network and highlights lines with Cisco DWDM and Open DWDM transmission systems. Figure 13: Map of CESNET2 network where CISCO lines are marked blue and lines lighted by open systems are in red First scenario used a loop over production lines equipped with Cisco ONS 15454 MSTP of total length of 1063 km by connecting Praha-Hradec Králové-Olomouc-Ostrava-Olomouc-Hradec Králové-Praha cities. Total overcame attenuation was 276 db with two spans of 22 db attenuation. Overall performance of the 100G system was measured by embedded AFEC performance and BERT running on one of multiplexed 10G channels. The 100G system was running smoothly with 33e+9 corrected errors in AFEC each 15 minutes. Second scenario had shorter total distance of 778 km, but 100G signal was sent over same line as a photonic service of precise atomic clock comparison (both 100G and photonic services were transmitted in the same fiber) [3]. The ring Praha-Hradec Králové-Olomouc-Brno-Praha (see Figure 13) had 208 db of attenuation and included also three spans with attenuation of 22 db. The 100G system was running smoothly with just 4e+9 corrected errors in AFEC each 15 minutes. During the test the photonic service was in operation on the line Praha-Brno with no evidence of influence on each other. Third scenario was conducted over combination of Cisco DWDM and Open DWDM transmission systems on line Praha-Brno-Wien-Brno-Praha with concurrent transmission of photonic service of atomic clock comparison. The total length of line was 1056 km with 285 db of attenuation and 4 spans of 22dB plus another four critical spans of 30, 31, 32 and 34 db. AFEC reported 303e+9 corrected errors and BERT measured uncorrected bit errors of 25e+3 in 15 minutes. The main reason behind this effect is probably the reduction of signal quality at four critically long spans. Fourth scenario was set over single fiber bidirectional transmission lines. The ring Praha-Plzeň-Cheb- Most-Ústí nad Labem-Praha included part of Cisco DWDM and part of Open DWDM transmission systems in total length of 655 km and 180 db of attenuation. The ring had three critical spans of 35.5, 34, 35.5 db. AFEC corrected 118e+9 errors over 15 minutes.

5 Conclusion 100G coherent transmission system is resilient to many network impairments as nonlinearities, filtration and inter-channel interference. Coherent systems can work over compensated networks with no known influence from type of dispersion compensation. During comparison of DCF, FBG and uncompensated scenarios, no impact, but more attenuation according to DCF length, has been observed. FBG due to their lower insertion loss, nonlinear effects and delay should be preferred for chromatic dispersion compensation. Therefore new coherent systems can be deployed into already working 10G NRZ transmission systems as long as network will remain dispersion compensated. ALU 100G Photonic Service Switch Platform verified to work with Cisco and CzechLight transmission systems in multivendor environment. The ALU 100G also proved to work over single fiber bidirectional transmission lines and in parallel with the Photonic Services of atomic clocks comparison. References [1] Essiambre, R.-J.; Kramer, G.; Winzer, P.J.; Foschini, G.J.; Goebel, B.;, "Capacity Limits of Optical Fiber Networks," Lightwave Technology, Journal of, vol.28, no.4, pp.662-701, Feb.15, 2010 [2] Roberts, K.; O'Sullivan, M.; Kuang-Tsan Wu; Han Sun; Awadalla, A.; Krause, D.J.; Laperle, C.;, "Performance of Dual-Polarization QPSK for Optical Transport Systems," Lightwave Technology, Journal of, vol.27, no.16, pp.3546-3559, Aug.15, 2009 [3] V. Smotlacha, A. Kuna, W. Mache: Time Transfer Using Fiber Links, EFTF 2010, Noordwijk, The Netherlands, 2010 Biographies Pavel Škoda joined in 2005 the project Components for high transmission rate all-optical networks at the Institute of Photonics and Electronics at the Academy of Sciences of Czech Republic. He graduated in 2008 at the faculty of electrical engineering of the Czech Technical University in Praha. After graduation Pavel went to the Tyndall National Institute in Ireland to study the dynamics of mutually coupled laser system. Since 2009 he has been working at Optical networks activity in CESNET z.s.p.o. in Praha. In 2010 Pavel started the Ph.D. study at the Czech Technical University in Praha. Jan Radil received the M.Sc. and Ph.D. degrees in electrical engineering from the Czech Technical University, Praha, in 1996 and 2004, respectively. Jan joined the Research and Development Department, CESNET, Praha, in 1999, where he is responsible for optical networking and the development of the next generation of the Czech research and educational network. Jan participated in the EU projects SEEFIRE, Porta Optica Study, Phosphorus, GN2 and GN3. Josef Vojtěch received with honors M.Sc. degree in electrical engineering, B.Sc. degree in pedagogy and Ph.D. degree from the Czech Technical University, Praha, in 2001, 2003 and 2009 respectively. Since 2003, he has been with Optical networks project of Research and Development Department of CESNET, a.l.e., where he is responsible for development of open family of photonic devices. He is a member of IEEE and OSA and holds 7 patents and utility models with others pending. Miloslav Hůla joined Research and Development Department of CESNET, a.l.e., in 2007, where he is active in applied research in the area of photonic networking. He received the M.Sc. degree in Electrical Engineering from the Czech Technical University, Praha, in 2008.