Travel & Places

Field trial on 107-Gbit/s transmission with pure electrical timedivision multiplex over 160 km

Field trial on 107-Gbit/s transmission with pure electrical timedivision multiplex over 160 km
of 5
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
  Field trial on 107-Gbit/s transmission with pure electrical time-division multiplex over 160 km Rainer H. Derksen 1 , Colja Schubert 2 , Sander Jansen 3 , Xiang Zhou 4 , Martin Birk  4   1  Siemens Networks GmbH & Co. KG, MN PG NT CT 1, Otto-Hahn-Ring 6, D-81730 München, Germany, Fon: +49-89-636-43460, Fax: +49-89-636-45814, Email:, Internet:  2  Fraunhofer Institute for Telecommunications, Heinrich-Hertz-Institut, Berlin, Germany 3  COBRA Institute, Eindhoven University of Technology, The Netherlands, now with KDDI R&D Laboratories, Saitama, Japan 4  AT&T Labs - Research, Middletown, NJ 07748, USA Abstract Electrical time-division multiplexing is the most cost-effective way to increase the transmission capacity of a single wavelength channel. Given the increasing importance of Ethernet-based data traffic there has recently been a strong interest to investigate the feasibility of 107-Gbit/s transmission with pure electrical time-division multiplex as a basis for a future 100-Gbit/s Ethernet standard. The main challenge to realize 100-Gbit/s Ethernet is the development of cost-effective systems. Systems based on pure electrical time-division mulitplex that furthermore are compatible with today’s channel spacing in the optical band used by wavelength-division multiplexing are therefore preferable. Meanwhile, for the transmission of 100-Gbit/s Ethernet different modulation formats have been examined, such as on-off keying, duobinary coding and differential quadrature-phase shift keying. The most straightfor-ward implementation for 100-Gbit/s Ethernet that requires the least optical and electrical components is the on-off keying modulation format. Several investigations on 100-Gbit/s Ethernet have been published using this modulation format, however, so far no transmission experiment with pure electrical time-division multiplex both on the transmitting side and the receiving side has been reported. Here we report on the first 107-Gbit/s transmission with on-off keying and pure electrical time-division multi-plex in the transmitter and receiver - namely over 160 km installed fiber in the network of a major American carrier. Besides the presentation of the achieved results, the still existing bottle necks - above all caused by the use of components which srcinally had been developed for operation with 40 Gbit/s or 80 Gbit/s only - are discussed. From this the resulting requirements for further research work are deduced. Furthermore we will show the measures applied to enable the transmission of 107-Gbit/s on-off keyed signals within a bandwidth of 100 GHz, required for the standard ITU-T channel spacing.  Deutsche Kurzfassung - Feldversuch zur 107-Gbit/s-Übertragung mit rein e-lektrischem Zeitmultiplex über 160 km  Elektrisches Zeitmultiplex ist die kostengünstigste Art, die Übertragungskapazität auf einem Wellenlän-genkanal zu erhöhen. Vor dem Hintergrund zunehmender Bedeutung von Ethernet-basiertem Datenver-kehr entstand in letzter Zeit großes Interesse daran, die Machbarkeit von 107-Gbit/s-Übertragung mit rein elektrischem Zeitmultiplex als Grundlage für einen zukünftigen 100-Gbit/s-Ethernet-Standard zu untersu-chen. Die größte Herausforderung bei der Realisierung von 100-Gbit/s-Ethernet ist die Entwicklung kos-tengünstiger Systeme. Auf rein elektrischem Zeitmultiplex basierende Systeme, die zudem noch kompatibel sind zu dem derzeitigen Kanalraster des im Wellenlängenmultiplex genutzten optischen Bandes, stellen deshalb eine bevorzugte Lösung dar. Für die Übertragung im 100-Gbit/s-Ethernet wurden mittlerweile verschiedene Modulationsarten unter-sucht, wie z. B. binäre Intensitätsmodulation, Duobinärkodierung und differentielle Quadratur-Phasenumtastmodula-tion. Das einfachste Verfahren für 100-Gbit/s-Ethernet, das die wenigsten optischen und elektrischen Komponenten benötigt, ist die binäre Intensitätsmodulation. Zwar sind bereits einige Un-tersuchungen zu 100-Gbit/s-Ethernet mit diesem Modulationsverfahren erschienen, jedoch wurde bislang  noch nicht von einem Übertragungsversuch mit rein elektrischem Zeitmultiplex sowohl auf Sender- als auch Empfängerseite berichtet.  Hier nun berichten wir über die erste 107-Gbit/s-Übertragung mit binärer Intensitätsmodulation und rein elektrischem Zeitmultiplex - und zwar über 160 km verlegte Standardglasfaser im Netz eines großen ame-rikanischen Netzbetreibers. Neben der Präsentation der erzielten Ergebnisse werden die noch vorhande-nen Engpässe - vor allem durch die Verwendung von Komponenten, die ursprünglich nur für den Betrieb mit 40 Gbit/s oder 80 Gbit/s entwickelt wurden - diskutiert und die daraus resultierenden Erfordernisse für weiterführende Forschungsarbeiten abgeleitet. Außerdem werden die in diesem Feldversuch ergriffenen  Maßnahmen dargestellt, die es ermöglichten, das 107-Gbit/s-Signal innerhalb der dem derzeitigen Kanal-raster entsprechenden Bandbreite von 100 GHz zu übertragen. 1 Introduction Electrical time division multiplexing (ETDM) is the most cost-effective way to increase the capacity of a single wavelength channel. Given the increasing im-portance of Ethernet-based data traffic there has re-cently been a strong interest to investigate the feasi-bility of 107-Gbit/s ETDM transmission as a basis for a future 100-Gbit/s Ethernet (100GbE) standard [1]. The main challenge to realize 100GbE is the de-velopment of cost-effective systems. Full-ETDM based systems that are compatible with today’s wave-length division multiplexing (WDM) fiber infrastruc-ture are therefore preferable. Recently many research groups showed interest in 100GbE and different modulation formats have been proposed, such as OOK [2, 3], duobinary [4] and DQPSK [5]. The most straightforward implementa-tion for 100GbE that requires the least optical and electrical components is the OOK modulation for-mat. However, so far no 107-Gbit/s full ETDM transmission experiment using OOK has been re-ported. In this paper we show for the first time full-ETDM 107 Gbit/s transmission. Transmission is reported over 160 km of installed fiber. The optical bandwidth of the modulator used in this experiment is low (about 30 GHz). Therefore, in this experiment an FIR filter had to be used to compensate for the imperfec-tions of the modulator. The FIR filter was srcinally designed for operation at 40 Gbit/s and therefore had a small free spectral range of 91 GHz [6]. This puts stringent requirements on the available optical band-width for the data signal. By using the vestigial side-band (VSB) modulation format, the optical bandwidth of the 107 Gbit/s signal is significantly reduced. As a result, the signal becomes as well extremely tolerant against narrowband optical filtering. Although the VSB modulation format requires a high OSNR, it en-ables cascaded filtering of the 107-Gbit/s data signal with optical filters having a 3-dB bandwidth of 0.74 nm (92.5 GHz). 2 Experimental setup The experimental setup is depicted in Fig. 1. Fig. 1  Experimental setup, map and eye diagrams for (a) electric back-to-back, (b) after optical modu-lator and (c) after VSB and FIR filter. The data and inverted data outputs of a 53.5-Gbit/s pulse pattern generator (PPG, provided by SHF tech-nologies) are fed to a SiGe 2:1 multiplexer (SHF 408). For the experiment a pseudo-random bit se-quence (PRBS) with length 2 7 -1 is used. The PRBS length is mainly limited by the electrical bandwidth of the receiver. In order to decorrelate the two tribu-taries, the inverted data output is delayed by 16 bits  with respect to the data output. The balanced output of the 107-Gbit/s multiplexer provides a peak-to-peak voltage of about 800 mV and is directly fed into a dual-drive Mach-Zehnder modulator with a 3-dB bandwidth of about 30 GHz and V π  = 3 V at 1.5 GHz. A distributed feedback (DFB) laser provides a con-tinuous wave at 1551.64 nm for modulation. Subse-quently, the signal is VSB filtered with a tunable bandpass filter (BPF). The center wavelength of the VSB filter is located at 1550.7 nm and the 3-dB and 20-dB filter bandwidths are 1.0 nm and 2.78 nm, re-spectively. The optical equalizer used is a finite impulse re-sponse (FIR) lattice filter integrated on a SiON pla-nar lightwave circuit. The filter is a six-tap feed for-ward filter, has a free spectral range of 91 GHz and is described in more detail in [6, 7]. Comparing the eye diagrams depicted in Fig. 1 it can be seen that the combination of VSB and FIR filter significantly im-proves the eye opening. However, note that the eye diagrams shown are for 100 Gbit/s, since the preci-sion timebase of the oscilloscope introduced a large  jitter when used at the 13.375-GHz base rate required for measurements at 107 Gbit/s. Before transmission a dispersion compensating fiber (DCF) module with –40 ps/nm is used as pre-compensation. The transmission link consists of two spans of 80-km fiber installed between Asbury Park and Little Egg Harbor (as depicted in the map of Fig. 1). These fibers have an average span loss of 19.7 dB and a dispersion of approximately 4.3 ps/nm/km at 1551.64 nm. After each span a DCM is used to com-pensate the chromatic dispersion. In front of the re-ceiver the residual dispersion is optimized with addi-tional fiber to obtain the best bit-error ratio (BER) performance. The input power into the transmission fiber and DCMs is 7 dBm and 0 dBm, respectively. The 107-Gbit/s signal is detected with a high-speed photodiode (provided by u2t photonics). Subse-quently, the electrical 107-Gbit/s signal is fed into the integrated ETDM receiver chip [2], which com-prised a 1:2-demultiplexer (DEMUX) and the clock and data recovery (CDR). An external voltage con-trolled oscillator (VCO, provided by Agilent Tech-nologies) is used, which is driven by the phase-detector signal generated in the receiver chip. The demultiplexed 53.5-Gbit/s tributaries are subse-quently detected with a BER tester (provided by SHF technologies) synchronized to the recovered 53.5-GHz clock signal from the receiver chip. 3 Experimental results Table 1 summarizes the observed Q-factor and opti-cal signal-to-noise ratio (OSNR) before and after transmission over the 160-km installed fiber link. Af-ter transmission the Q-factors are about 0.5 dB above the correction limit of a concatenated FEC code with a 7% overhead (Q-factor 9.0 dB). Both the electrical and optical signals are severely band-limited and the amplitude and phase response of the optical equalizer and of the VSB filter have a strong influence on the transmission performance. Q-factor Tributary 1 / 2 OSNR Back-to-back 10.4 dB / 10.5 dB 42 dB After 160 km 9.6 dB / 9.7 dB 35.6 dB Table 1 Performance before and after transmission.   . The amplitude response of the FIR, VSB and BP filter is depicted in Fig. 2a, as well as the cumulative amplitude response. An advantage of the VSB filtered signal is that it is more resistant to narrowband opti-cal filtering [8], although it requires a higher OSNR. We measured the resistance of the 107-Gbit/s signal to narrowband optical filtering using a flat-top nar-rowband BPF, with a 3-dB bandwidth of 0.74 nm and a 20-dB bandwidth of 0.89 nm (depicted in Fig. 2a). Fig. 2b depicts the optical spectrum of the 107-Gbit/s NRZ signal as well as the VSB signal before and after the narrowband filter. a) Filter curves. b)   107-Gbit/s NRZ and VSB with and without BPF Fig. 2  Optical spectra (0.01 nm resolution band-width). After filtering the 107-Gbit/s signal with the narrow-band BPF, a negligible increase of the BER was seen. Even for two (identical) cascaded narrowband BPFs no noticeable penalty with respect to back-to-back was observed. Hence it can be concluded that the 107-Gbit/s VSB modulated data signal is sufficiently  tolerant towards narrowband optical filtering to be used within an existing 100-GHz WDM infrastruc-ture. In order to avoid possible penalties resulting from linear and nonlinear crosstalk from neighboring channels 200-GHz spaced 107 Gbit/s channels can be interleaved with 40/10 Gbit/s traffic. The VSB filter plays a crucial role in increasing the tolerance of the 107-Gbit/s optical signal with re-spect to narrowband optical filtering. Fig. 3a and 3b show the simulated eye diagrams without and with narrowband BPF, respectively. a) simulation without the VSB filter b) simulation with the VSB filter c) experiment with VSB filter Fig. 3  Eye diagrams at 107 Gbit/s after filtering with the narrowband BPF. Simulated is an ideal 107-Gbit/s optically modulated signal after being filtered through the narrowband BPF. The center frequency of the narrowband BPF is in both configurations optimized to obtain the small-est eye opening penalty (EOP). The obtained EOP with and without VSB filter is 2.3 dB and 3.8 dB, re-spectively. Therefore it can be concluded that the re-sistance to narrowband optical filtering is signifi-cantly increased through the VSB filter. We conjec-ture that the main reason the filtering penalty was not observed experimentally is because the modulator impairments in combination with the small FSR of the FIR filter already reduce the effective optical bandwidth of the signal. Fig. 3c shows the obtained experimental eye diagram of the 107-Gbit/s VSB signal after the narrowband BPF. Please note that ad-ditional jitter is present in this eye diagram, caused by the precision timebase of the oscilloscope. As well in simulation as in experiment, before and after narrowband filtering a similar eye diagram is ob-served. 4 Conclusions and discussion We demonstrated for the first time 107-Gbit/s full-ETDM transmission over 160 km field installed fi-ber. Furthermore, we showed a high tolerance to-wards narrowband optical filtering by using 107-Gbit/s vestigial sideband modulation. Nevertheless there are still some bottle necks. On the receiver side the integrated receiver chip after the photo diode was operated at its limit since it was srcinally designed for 86-Gbit/s operation. But this circuit can be designed for nominal operation at 107 Gbit/s with some margin and including an on-chip voltage-controlled oscillator. The most severe limitation is on the transmitter side, especially the modulator. Modulators available today have not enough bandwidth and need too high driving voltages. Therefore, modulators with bandwidths of at least 80 GHz have to be developed with required in-put voltages which can be delivered by state-of-the-art integrated high-speed circuits. These are contra-dictory requirements. To mitigate this problem, also driver circuits with higher output voltages should be developed. In our experiment we could achieve the transmission only by use of the mentioned optical filter to com-pensate for the low performance of the modulator. Though this gave us the additional advantage to fit the spectrum into the current channel spacing of installed wave-division multiplex systems, in the long run it is undesirable to use this additional device. Then, of course, the possibility of fitting the spectrum to the existing channel spacing gets lost. Therefore, also alternative solutions should be inves-tigated, as e. g. using DQPSK, which has the advan-tage that bit rate dependent effects on the fiber are mitigated since the transmission of 100 Gbit/s is ac-complished with a symbol rate of only 50 GBaud. Additionally due to this lower symbol rate, the spec-trum of the signal can be fitted more easily into the existing channel spacing. On the other hand more complicated and costly transmitter and receiver de-vices are required. Both ways should be followed further. Possibly both solutions might co-exist in the end - OOK as a sim-ple, straightforward and cost-effective method for short reach applications (as currently discussed in the IEEE High-Speed Study Group), DQPSK for long-  haul transmission, where the more expensive trans-mitters and receivers are justified by the longer reach. Another interesting candidate for 100-Gbit/s trans-mission is the use of polarization multiplexed QPSK with coherent detection, requiring a symbol rate of only 25 GBaud, thereby encountering only little problems in the optical domain and easily fitting into the existing channel spacing. The problems are shifted to the electrical domain, analog-to-digital converters are required with a performance not yet achievable and a complex signal processing in the digital domain. In consequence, the development of such a solution has a high risk and requires a large effort. But in case this solution is viable one gets rid of most of the problems srcinating from the use of high symbol rates. Summing up, all three discussed methods should be investigated further. 5 Acknowledgements We thank Dr. Walthes from Infineon Technologies for the support under the MultiTeraNet program. We also thank u2t photonics AG, SHF Communications Technologies AG, and Agilent Technologies for pro-viding some of the measurement equipment. The work was partly funded by the BMBF project EI-BONE under the contract number 01 BP 560. 6 Literature [1] M. Duelk et al., ECOC 2005, Tu 3.1.2 [2] C. Schubert et al., ECOC 2006, Tu1.5.5 [3] K. Schuh et al., ECOC 2006, We3.P.124 [4] P. Winzer et al., JLT, 2006, vol. 24, pp. 3107-3113 [5] M. Daikoku et al., OFC 2006, PDP 36 [6] F. Horst et al., PTL, 2005, vol. 15, pp. 1570-1572 [7] M. Bohn et al., OFC 2005, OWO4 [8] S. Bigo, JSTQE, 2004, vol. 10, 329-340
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!