- Email: [email protected]
s carriers in a flexgrid super-channel arrangement
Optical Switching and Networking 19 (2016) 155–164
Contents lists available at ScienceDirect
Optical Switching and Networking journal homepage: www.elsevier.com/locate/osn
Reconfigurable DP-16QAM/QPSK transponders for the transmission of coherent 200 Gb/s carriers in a flexgrid super-channel arrangement Claudio Colombo Alcatel-Lucent, Italy
a r t i c l e i n f o
abstract
Article history: Received 31 October 2014 Received in revised form 1 September 2015 Accepted 2 September 2015 Available online 14 September 2015
The next generation of Dense Wavelength Division Multiplexing (DWDM) networks are likely to use Flexgrid arrangement, providing operators with additional flexibility when assigning spectrum compared to traditional DWDM networks using the 50 GHz ITU grid. Flexgrid breaks the spectrum up into small (typically 12.5 GHz) slots, but its key feature is that contiguous slots can be joined together to form arbitrary sized blocks of spectrum. This additional flexibility will allow faster transponders that utilize high spectral efficiency modulation techniques, like multi-level m-QAM schemes. From the use of these new spectrum efficient modulation formats and finer control over spectrum allocations, a key benefit that Flexgrid offers to the network operators is that their DWDM networks can carry more traffic with optimized spectral efficiency. High speed technology became essential for realizing greater network capacity and enabling network operators to meet the increasing bandwidth challenges from new generation superfast devices, services and applications. Coherent technologies beyond 100G are now focusing on higher level modulation formats and multiple sub-carriers/channels, using super-channels to achieve Terabit transmission. & 2015 Elsevier Ltd All rights reserved.
Keywords: Flexgrid Super-channel Coherent detection DWDM Spectral efficiency
1. Introduction Service providers around the world are moving quickly to accelerate their networks, making the jump from transport speeds of 10G and 40G to barrier breaking 100G and beyond. With the introduction of the first commercial 400G electro-optical chips, the recent technological evolution is already opening the way for future high-speed optical networks, allowing the optimal tradeoff between spectral density and optical reach per deployed line system. Moreover the benefits of Flexgrid arrangement become readily evident when coupled with next-generation DSP, supporting for higher baud-rate transmission. Improving flexibility while increasing capacity of the photonics line allows giving service providers faster access to the latest advances in optical and silicon technologies on their switched optical http://dx.doi.org/10.1016/j.osn.2015.09.001 1573-4277/& 2015 Elsevier Ltd All rights reserved.
networks, and the potential evolution for the optical platforms it supports. From the beginning of 2013 many experimental trials have been proposed to demonstrate an evident technological advantage in high capacity optical transmission by different manufacturers, through the networks of the main International Global operators. Some of the more interesting results will be presented and discussed in the paper, especially concerning the linear properties of transmission.
2. Key innovation technology The push for 400G consolidates when 100G transport is an increasing industry success story. Hundreds of commercially deployed 100G systems across the globe and continued bandwidth growth pressures motivate the
156
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
industry increased focus on 400G and 1 Tb next generation of higher transmission systems to drive down cost of transported bit. Steady advance in optical and electronic transceiver technology supporting advanced modulation and spectral shaping techniques motivate the rate increasing. 400G data rate is a natural next step gives the current evolution of datacom and telecom transport interface speeds and implementation complexity. 400G and beyond R&D activities occurring now in order to build the technology innovations and sustainable ecosystem for a commercial future. Various ways to realize 400G transmission rate have been proposed:
increasing the symbol rate, increasing the number of bonded channels, increasing the modulation levels, or their combination For long haul distances, bonding a number of established PM-QPSK channels via WDM techniques can result in high spectrum efficiency and lower bandwidth requirements. For metro distances, higher‐level QAM formats can be used to achieve spectral efficiency that is greater than that of PM‐QPSK. Dual‐Carrier PM‐16QAM coding on 75– 100 GHz grid is an attractive solution to realize 400G line rates while at the same time relaxing the high speed bandwidth requirements of the electronic circuits and optoelectronic components. The coherent 200G/16QAM optical solution can provide many customer benefits: avoid saturating photonic lines and costly overlays, propagate with 10G NRZ, 40G DP-BPSK and 100G DP-QPSK, propagate over other vendor’s photonic line ( Fig. 1). Compared to today’s 100G chips, the new 400G Photonic Service Engine (PSE) designed by Alcatel-Lucent and tested in the described trial links delivers: 4 times the line rate; more than 2.6 times the spectral efficiency; more than 4 times the density; 33% lower power consumption and 25% better tolerance to fiber impairments. As a result, the device reaches the current technological limits of the market’s silicon for integrated digital signal processing in high-speed telecom applications ( Fig. 2). According to the key enable technologies to reach the high-speed fiber optics performances
DWDM solution to reach multiterabit/s capacity, with a
super-channel configuration in an optical multi carriers arrangement Reconfigurable Nodes and Flexible-grid operation Modulation format, advanced DSP, soft-decoding FEC and Coherent detection,
the main functionalities achieved during all the field-trial activities have been focused on:
High-speed 400G transmission over two aggregated sub-carriers 16QAM modulation scheme.
Integrated wave-shaping, pre-compensation and enhanced
optical signal to noise ratio (OSNR) performance, enabling very high tolerance to fiber non-linearities. Advanced algorithms enabling SW-programmable adaptive modulation format and bit-rate reconfiguration on the same optical channel including PDM-BSPK, PDM-QPSK and PDM-16QAM to optimize reach and capacity. Spectral efficiency improvement, demonstrating the benefits of “Flexible-grid Optical Networks” combining the best relation between distance and speed, according to the ITU-T Rec. G.694-1 [1]. Support for 50 GHz spacing at 400G for full compatibility with reconfigurable optical add-drop multiplexer (ROADM) networks.
With the optical technology available today, service providers must choose a specific modulation format for each signal speed. The 400G PSE forms the basis for software-defined flexible bit-rate and adaptive modulation line interfaces. A transponder based on this ASIC generation can generate a number of spectrally engineered modulation formats. This gives service providers the flexibility to choose the optimal modulation format for their distance and spectral efficiency requirements.
3. Capacity and bandwidth efficiency Bandwidth efficiency and increased role of digital signal processing in the optical layer have been the key elements to allow the multi-terabit transmission over the traditional channel planning and the flex-grid arrangement, recognized as the revolutionary technique for high capacity future systems. Considering, at the beginning, the limit for a linear channel with an additional Gaussian white noise distribution, the maximum capacity Cmax (maximum spectral efficiency) can be expressed as a function of channel width CW by the relation: C max ¼ CW log2 ð1 þSNRÞ In Fig. 3 the theoretical trend of the channel capacity as a function of OSNR OSNR ¼ CW=12:5 SNR is described, for different channel widths equal to the DWDM grid channel spacing. The theoretical OSNR degradation can be evaluated for a channel width of 37.5 GHz respect to the fixed 50 GHz WDM grid: 1 dB for a channel capacity of 100G, 2.5 dB for a capacity of 200G, 7 dB for a capacity of 400G. Software re-configurable QPSK/16QAM transponders guarantee excellent optical performances for both the adaptive rate (100 Gb/s and 200 Gb/s) in terms of OSNR and Q-factor. Figs. 4 and 5 show some significant measurements performed with standard production transponders. At the relative SD-FEC thresholds [BER ¼2.1 10 2], the deviations of measured OSNR respect to the theoretical
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
157
Fig. 4. Pre-FEC BER vs. OSNR measurements.
Fig. 1. QPSK and 16-QAM constellation diagrams.
Fig. 5. Q-factor vs. OSNR measurements.
Fig. 2. Evolve to 100G and beyond.
Fig. 3. Channel capacity vs. OSNR.
curves are really close: OSNRmeasured OSNRtheoretical ¼ 12:5 dB 11:0 dB ¼ 1:5 dB for 100G QPSK 20:5 dB 17:5 dB ¼ 3:0 dB for 200G16QAM The physics tradeoff of using 16QAM modulation vs. QPSK is that maximum reach decreases as bits/symbol
increase. For the 16QAM demonstrations, distances achieved are in the regional range of 400–600 km. Therefore, significant advances in FEC technology would be required in order to extend the 16QAM-based systems into long-haul and ultra-long-haul applications. Of course, the same OSNR deviations can be evaluated for the Q-factor limit of 6 dB, corresponding to the relative pre-FEC BER threshold of 2.1 10 2 for both the systems. The experimental links at high speed/capacity (41 Tb/s) have been demonstrated adjusting a super-channel of 7 200G carriers 16QAM modulated, with an optimized digital pulse shaping and operating with a spectral efficiency up to 5.7 bits/s/Hz. The 1.4 Tb/s super-channel is set up with seven 200G sub-channels generated from the software re-configurable transponders. In comparison with standard 50 GHz spacing, Fig. 6 shows the super-channel spectrum with sub-channel spacing at 50 GHz and 35 GHz respectively. A raised-cosine filter has been used for TX pulse shaping, reducing the crosstalk and the inter-symbol interference [2]. The spectral reduction from 350 GHz to 245 GHz represents an effective bandwidth occupancy improvement of about 30% (Fig. 7). In the described field trial, the 1.4 Tb/s Alien Superchannel propagates alongside the existing 40G and 100G native wavelengths over the trial link. Fig. 8 shows the measured optical spectrum of the native wavelengths and 1.4 Tb/s Alien Superchannel after 410 km transmission. The
158
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 10. C-band capacities in split spectrum and gridless arrangement.
Fig. 6. 1.4 Tb/s super-channel spectrum.
Fig. 7. Spectral reduction in flexgrid arrangement.
Fig. 8. Optical wavelengths.
spectrum
of
the
1.4 Tb/s
and
native
40G/100G
figure is indicative only and directly downloaded from an optical spectrum analyzer, so the bad quality depends from the instrument imaging processing. Fig. 9 shows the pre-FEC BER performance of the Alien Superchannel over 410 km for sub-channel spacing at 50 GHz, 40 GHz, 37.5 GHz, 36 GHz and 35 GHz respectively. All the seven 200G sub-channels achieved pre-FEC BER of better than 2 10 2, which gives error free performance after FEC. At 50 GHz spacing, the Alien Superchannel has the lowest pre-FEC BER as expected. As the channel spacing narrows, the pre-FEC BER degrades due to the increasing crosstalk between the adjacent channels. Higher order modulation will certainly be a useful tool to allow service providers to optimize the total spectral efficiency for certain routes. But the penalty for this is reduced reach. Unfortunately, higher order modulation on its own is not a solution in moving to Terabit capacities, and there is no single correct modulation technique for any given route. Once again, this highlights the need for a multi-carrier super-channel implementation with adaptive coherent modulation so that service providers can optimize the combination of reach and spectral efficiency without having to order multiple part numbers from their system supplier [3]. Fig. 10 illustrates the total transmitted capacity on the whole C-band for a selection of modulation schemes, in a fixed split spectrum condition and in a gridless arrangement. Again, for all the considered modulation schemes an evident advantage of 33.33% is demonstrated in a gridless configuration with respect to the standard fixed split spectrum arrangement at n 50 GHz. Fig. 11 shows the normalized reach (assuming the QPSK application as standard reference) for the different schemes: modulation techniques likes 16QAM may be limited to regional network use. This comparison is illustrative only: not all modulation types shown are practical, and other modulation types may be available in final products.
4. Flexgrid arrangement
Fig. 9. Performance of the Alien Superchannel over 410 km.
In the traditional approach to WDM, the optical transmission spectrum is carved into a number of nonoverlapping wavelength (or frequency) bands with spectrum spacing of 50 GHz (or 100 GHz), as specified by ITU-T standards, and each wavelength supports a single communication channel operating at a certain fixed rate
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 11. Normalized reach for different modulation formats.
(e.g., 10 Gb/s, 40 Gb/s, or 100 Gb/s). This is referred to as the fixed grid. Recent trends in optical transmission technologies such as coherent detection and digital signal processing are opening new horizons for novel flexible transmission systems and optical devices such as bandwidth-variable Optical Cross-connects or bandwidthvariable ROADMs. A more practical spectrum management for optical networks is the so-called flexible grid, in which the spectrum grid granularity is much finer than the ITU-T grid and a lightpath’s spectrum can span across several frequency slices. [4] Key technology elements to achieve flexible line are:
Flexgrid, utilizing reduced spacing in order to increase spectral efficiency and introduce new level of flexibility
Flexible wave-shaping, enabling a further increment in network capacity
Coherent technology, pick and choose modulation depending on networking need
Programmable capacity Resiliency and control
Comparison of fixed WDM and flexible grid approaches has been evaluated in terms of spectral efficiency improvement and spectral distribution optimization for future “elastic” or rate-variable optical networks. Fig. 12 compares the spectral efficiency for the different 200G sub-carriers spacing and the corresponding spectral efficiency improvement percentage respect to the fixed 50 GHz channel spacing solution. With the 35 GHz channel spacing limit, the spectral efficiency grows up to 6 bit/s/Hz with an improvement of more than 40%. A further comparison of fixed WDM and flexible grid approaches can be evaluated in terms of spectrum occupation and transmitted capacity. In the fixed WDM network, the total available spectrum of 4.4 THz is divided in 88 slices of 50 GHz each. In the flexible grid arrangement we consider a grid granularity of 6.25 GHz or 12.5 GHz, which leads to 704 or 352 frequency slices. The effective signal baud-rate of 28 Gbaud (symbol rate¼32.5 Gbaud and roll-off¼0.16) matches with both 31.25 GHz or 37.5 GHz spacing, corresponding to 5 slices of 6.25 GHz width or 3 slices of 12.5 GHz width, respectively. In both the fixed grid and flexible grid arrangement, we consider adaptive digital coherent transceivers supporting
159
polarization multiplexed BPSK, QPSK and n-QAM modulation formats, with n ¼8, 16, 32, 64, 128. The transceiver utilizes the most efficient modulation format compatible with the distance to be covered by the lightpath, with the associated bit-rate of each single traffic request. Introducing a flexgrid arrangement allows a significant improvement in terms of spectral efficiency, especially using a channel bandwidth of 5 6.25 GHz (as demonstrated in Fig. 13) where, for a bit-rate of 200 Gb/s, a total efficiency of 6.4 bit/s/Hz is reached with a 16QAM modulation format, comparing with 4 bit/s/Hz using a 50 GHz WDM grid. A possible future implementation of 400G singlecarrier with an available 128QAM scheme in a channel bandwidth of 5 6.25 GHz per carrier could reach a spectral efficiency of about 13 bit/s/Hz, instead of 8 bit/s/ Hz using a 50 GHz grid. The spectrum savings (or spectral efficiency increasing) in flexible grid scenario with respect to the fixed WDM grid are uniquely motivated to reducing the frequency slice width. The slight reductions in the overall spectrum occupied are obtained due to finer capacity granularity, which allows for better adaptation of the number of transceivers to the traffic volume to be served, thus reducing the amount of unused capacity on the installed devices and consequently leading to more efficient spectrum utilization [5,6]. The same results can be seen in terms of total transmitted capacity in the optical bandwidth. Current fiber usage is limited by the fixed 50 GHz ITU-T grid (Fig. 14): 88 100G channels ðPDM QPSK
50GHzÞ-8:8 Tbps
44 400G channels ð2 PDM 16QAM50GHzÞ -17:6 Tbps Flex grid brings improved spectrum usage. Actual grid can be selected to match the spectrum of the signal, with frequency slots of nx12.5 GHz or n 6.26 GHz. Superchannels wider than 50 GHz are feasible: 116 100Gsuper chs:ðPDM QPSK
37:5 GHzÞ
-11:6Tbps 140 100Gsuper chs:ðPDM QPSK
31:25 GHzÞ
-14:0Tbps 58 400Gsuper chs:ð2 PDM 16QAM
75 GHzÞ
-23:2Tbps 70 400Gsuper chs:ð2 PDM 16QAM6
2:5 GHzÞ
-28:0Tbps The total transmitted capacity (described in Fig. 15 as a function of different bit-rate) has been evaluated over the complete C-band frequency window (4400 GHz, equivalent to 88chs at 50 GHz). The bit rate Br is a function of the number of bits/ symbol M (modulation factor) given by the modulation scheme, the symbol rate Sr and the number of (sub-) carriers SC: Br ¼ M Sr SC With higher level modulation formats, the number of bits per symbol is increased. E.g. for DP-m-QAM
160
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 12. Spectral efficiency comparison.
Fig. 13. Spectral efficiency comparison.
Fig. 14. Example of fixed and flexible grid channels distribution.
Fig. 16. Transmitted capacity using modulation schemes optimization.
multiple of the spectrum slice width (n ¼5BW¼31.25 GHz) better fit the network configurations required to achieve optimal spectrum assignment. In Fig. 16 we can see the total transmitted capacity for different client bit-rate optimized with the suitable DP-mQAM modulation format. Considering a bit-rate of 200 Gb/s per carrier (6.25 GHz spaced), a maximum capacity of 28 Tb/s could be transmitted in the total available bandwidth. We will first look at the gains from using Flexgrid when compared to inverse multiplexing. Inverse multiplexing is one technique to enable higher bitrate services to continue to use a fixed grid. In this approach, multiple lower bitrate channels are multiplexed together to form a higher aggregated bitrate. For example, a 400 Gbit/s signal could be transmitted using 4 100 Gbit/s sub-signals each of which could continue to fit within a 50 GHz channel. This approach does have the drawback that more spectrum is utilized than is absolutely necessary: using 4 100 Gbit/s channels uses 200 GHz of optical spectrum compared to just 75 GHz for transferring a native 400 Gbit/s signal. The efficiency improvement percentage for flexible spectrum over a point-to-point link is shown in Fig. 17, assuming a 50 GHz grid for fixed DWDM and super-channels with different Spectral Width (SW) for the applied modulation schemes. The efficiency gain of using native transponders compared to inverse multiplexing is easy to calculate numerically for a point-to-point link using the formula given in the following equation: Efficiency
Fig. 15. Total transmitted capacity in fixed and flexgrid arrangement.
modulation schemes, the modulation factor is given as M ¼ 2 log2 ðmÞ A finer grid granularity (nx6.25 GHz) combined with transceivers working at baud rates close to an integer
gain ¼
SW f ixed ðSW f lexgrid þ SW guardband Þ SW f lexgrid þ SW guardband
SW stands for the spectral width of the transponder in Flexgrid and Fixed grid scenarios. For the Flexgrid scenario, a small amount of spectrum (76.25 GHz) needs to be placed between signals to form a guard band. Of course, the super-channel implementation with a multi-carrier arrangement exhibits the best efficiency gain. The 400G solution (4 100G) or 1Tb (10 100G) operating in an optimized flexgrid bandwidth arrangement of n 12.5 GHz would reach the higher efficiency gain (up to 150%) respect to the fixed n 50 GHz spectral width. Parameters to increase the capacity are mainly the number of carriers, the modulation format and the baudrate. The represented curves in Fig. 18 justify and demonstrate the bit rate improvement per single carrier with the usage of higher baud-rate solution and the
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 17. Flex-grid efficiency for different rate and modulation formats.
Fig. 18. How to increase capacity.
Fig. 19. Normalized Capacity*Reach product.
high spectrum efficiency with higher order modulation schemes. The Capacity*Reach product (normalized at the achieved value with a QPSK modulation scheme) as a function of the total transmitted C-Band capacity is described in Fig. 19. The transmission of 80 200G carriers, 16QAM modulated, allows a total capacity of 16 Tb/s in the C-band, with a maximum reach equivalent to 40% of the nominal reach achieved with the 100G QPSK modulated carriers, performing the half of the total transmitted capacity. The required OSNR depends from the adopted modulation scheme and the system symbol rate. An example of
161
Fig. 20. Required OSNR for different modulation schemes.
Fig. 21. Spectral efficiency and total capacity for different 400G arrangements.
OSNR distribution for working baud-rate at 28 Gbaud and 56 Gbaud is calculated in Fig. 20 for some amplitude modulation schemes. The OSNR degradation between the relative symbol rate at the same modulation format is 3 dB. Today, available technologies beyond 100G using higher order constellation than QPSK have led to reach limitations, mainly as regards the maximum transmission reach. PM-16QAM technology 200G or 400G superchannel (2 200G sub-carrier) is often proposed by vendors, despite of 7 dB extra-OSNR required vs. 100G PM-QPSK, more sensitive to filtering caused by cascade of ROADMs, a limited 400 km maximum reach instead of a significant spectral efficiency improvement (from 2 to 5.33 bit/s/Hz). In Fig. 21 the spectral efficiency and the total transmitted capacity on the whole C-band have been calculated for different 400G multi-carriers spectral configurations, assuming a spectral bandwidth optimized with n 12.5 GHz slices for each modulation scheme. The implemented 2 200G (16QAM modulation scheme) in 75 GHz channel spacing performs the higher spectral efficiency (5.33 bit/s/Hz) and the maximum transmitted capacity (23.2 Tb/s) corresponding to 58 equivalent 400G channels in the total optical available bandwidth. In Figs. 22 and 23 the constellation diagrams for the DP-16QAM modulation scheme (both polarizations) are shown. Transmit wave-shaping compresses signals approaching the Nyquist limit. The optimization of the baud-rate and the filter width allows achieving better performances:
162
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 22. 16QAM constellation (in back-to-back configuration).
Fig. 23. 16QAM constellation (after transmission over 400 km G.652 fiber).
compressed signals on 37.5 GHz grid see similar performance as standard signals on 50 GHz grid (see Fig. 24). In Fig. 25 the transfer function of the WSS filter with a bandwidth of 75 GHz has been optimized for the transmission of a flexible super-channel of 2 200G carriers, 37.5 GHz spaced. The image has been downloaded directly from the optical spectrum analyzer.
5. OSNR penalties vs. sub-channel spacing To achieve optimized performance, the spectral shaping filter is optimized corresponding to different channel spacing. A digital raised-cosine pulse shaping has been implemented with a roll-off optimized for a minimum BER floor for each carrier spacing value. Computation of OSNR penalties has been evaluated for sub-channels spacing down to 35 GHz. Fig. 26 shows the OSNR penalty curve for a sub-channel, which is at the middle of the Alien SuperChannel and suffers most from both linear and nonlinear
crosstalk from its adjacent channels. Compared to 50 GHz spacing, the OSNR penalty is within 0.5 dB for sub-channel spacing down to 37.5 GHz. Beyond this the optical penalty increases steeply. At 35 GHz spacing, error free performance is achieved with no margin. For channel spacing at 37.5 GHz or wider we measured stable, long term error free performance. The raised-cosine pulse shaping and the roll-off have been optimized for a minimum BER floor for each carrier spacing value, as indicated in Fig. 27 where are shown the correspondent pre-FEC BER curves. The transmit (TX) DSP wave-shaper can shape the transmit spectrum into any given shape based on filter configuration. This capability allows service providers to make trade-offs between spectral efficiency and non-linearity tolerance. It can optimize the overall spectral content of the signal, offering a higher degree of tolerance to noise and minimize crosstalk effects [7]. The bandwidth reduction of the optical signal allows denser channel spacing and increases capacity when
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
163
37.5 GHZ 50 GHZ
Fig. 24. Spectral distribution for different signal shaping.
Fig. 27. Pre-FEC BER vs. OSNR for different carrier spacings. Fig. 25. WSS filter (BW ¼75 GHz).
Fig. 28. OSNR penalty vs. normalized channel spacing. Fig. 26. OSNR penalty for different channel spacings.
flexible grids are employed, introducing OSNR penalties. Fig. 28 shows a plot of sensitivity as a function of the normalized sub-channel spacing Δf/Sr when using an
optimized TX digital filter. When the sub-channel spacing is less than 1.1 times the symbol rate the performances become unacceptable in terms of OSNR penalties. In order to match the granularity of the flexible grid, the channel bandwidth BW must be an integer multiple n
164
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
6. Conclusions
Fig. 29. Transmission reaches for different modulations and channel spacing.
of the slice width. The choice of n impacts on the performance of the transceiver: the higher the Sr/BW ratio, the higher the efficiency of the spectrum utilization, but the lower the maximum transmission reach length that can be covered without need of electronic regeneration. Assuming a symbol rate Sr ¼28 Gbaud and considering the channel bandwidths BW1 ¼50 GHz, BW2 ¼37.5 GHz (3 12.5 GHz) and BW3 ¼31.25 GHz (5 6.25 GHz): Sr =BW 1 ¼ 0:56;
Sr =BW 2 ¼ 0:75;
Sr =BW 3 ¼ 0:9
Fig. 29 reports the maximum reaches for different modulation formats (at the same symbol rate of 28 Gbaud) and the considered ratios. For a 16-QAM scheme the relative distance reduction is little more than 15% for a BW¼37.5 GHz and about 20% for a BW¼31.25 GHz, respect to the conventional WDM grid of 50 GHz. The reduction in reach is mainly caused by additional crosstalk due to adjacent channel coherent interference, and is more pronounced with lower modulation formats. The optical pre-filtering increases the system spectral efficiency, allowing the allocation of a higher number of channels in the same bandwidth and maintaining the modulation format at a penalty cost of transmission performance due to the intersymbol interference caused by the filtering. Thus, pre-filtering systems should be designed considering a trade-off between spectral efficiency, performance and maximum reach. This is an important factor when we consider that the future of optical networks tends to flexible applications of bit rate, modulation formats and channel bandwidth. [8]
The next generation of Dense Wavelength Division Multiplexing (DWDM) networks are likely to use Flexgrid arrangement, providing operators with additional flexibility when assigning spectrum compared to traditional DWDM networks using the 50 GHz ITU grid. This additional flexibility will allow faster transponders that utilize high spectral efficiency modulation techniques, like multilevel m-QAM schemes. From the use of these new spectrum efficient modulation formats and finer control over spectrum allocations, a key benefit that Flexgrid offers to the network operators is that their DWDM networks can carry more traffic with optimized spectral efficiency. High speed technology became essential for realizing greater network capacity and enabling network operators to meet the increasing bandwidth challenges from new generation superfast devices, services and applications. Coherent technologies beyond 100G are now focusing on higher level modulation formats and multiple sub-carriers/channels, using super-channels to achieve Terabit transmission. Within the next years, 400G is set to become the dominant backbone technology in terms of capacity shipped, replacing longtime incumbent 10G and 40G optical lines. OTN switching, ASIC chip design, coherent detection and soft-decision FEC are among the best innovations that will make 400G a success. The demonstration of these implemented features was evident in the awareness of the high technology level inherent to the proposed solution.
References [1] ITU-T Recommendation G.694-1, Spectral grids for WDM applications: DWDM frequency grid. [2] Yu Rong Zhou, et al., 1.4 Tb Real-Time Alien Superchannel Transport Demonstration over 410 km Installed Fiber Link Using Software Reconfigurable DP-16QAM/QPSK, pdp OFC, 2014. [3] T. Xia, et al., High-capacity optical transport networks, IEEE Commun. Mag. 50 (11) (2012) 170–178. [4] O. Gerstel, et al., Elastic optical networking: a new dawn for the optical layer? IEEE Commun. Mag. 50 (2) (2012) s12–s20. [5] C. Rottondi, et al., Optical ring metro networks with flexible grid and distance-adaptive optical coherent transceivers, Bell Labs Tech. J. 18 (3) (2013) 95–110. [6] C. Rottondi, et al., Routing, modulation level, and spectrum assignment in optical metro ring networks using elastic transceivers, J. Opt. Commun. Netw. 5 (4) (2013) 305–315. [7] P. Winzer, et al., Penalties from In-Band Crosstalk for Advanced Optical Modulation Formats, in: Proceedings of the ECOC, 2011. [8] G. Bosco, et al. Investigation on the robustness of a Nyquist-WDM terabit superchannel to transmitter and receiver non-idealities, in: Proceedings of the ECOC, 2010.
Contents lists available at ScienceDirect
Optical Switching and Networking journal homepage: www.elsevier.com/locate/osn
Reconfigurable DP-16QAM/QPSK transponders for the transmission of coherent 200 Gb/s carriers in a flexgrid super-channel arrangement Claudio Colombo Alcatel-Lucent, Italy
a r t i c l e i n f o
abstract
Article history: Received 31 October 2014 Received in revised form 1 September 2015 Accepted 2 September 2015 Available online 14 September 2015
The next generation of Dense Wavelength Division Multiplexing (DWDM) networks are likely to use Flexgrid arrangement, providing operators with additional flexibility when assigning spectrum compared to traditional DWDM networks using the 50 GHz ITU grid. Flexgrid breaks the spectrum up into small (typically 12.5 GHz) slots, but its key feature is that contiguous slots can be joined together to form arbitrary sized blocks of spectrum. This additional flexibility will allow faster transponders that utilize high spectral efficiency modulation techniques, like multi-level m-QAM schemes. From the use of these new spectrum efficient modulation formats and finer control over spectrum allocations, a key benefit that Flexgrid offers to the network operators is that their DWDM networks can carry more traffic with optimized spectral efficiency. High speed technology became essential for realizing greater network capacity and enabling network operators to meet the increasing bandwidth challenges from new generation superfast devices, services and applications. Coherent technologies beyond 100G are now focusing on higher level modulation formats and multiple sub-carriers/channels, using super-channels to achieve Terabit transmission. & 2015 Elsevier Ltd All rights reserved.
Keywords: Flexgrid Super-channel Coherent detection DWDM Spectral efficiency
1. Introduction Service providers around the world are moving quickly to accelerate their networks, making the jump from transport speeds of 10G and 40G to barrier breaking 100G and beyond. With the introduction of the first commercial 400G electro-optical chips, the recent technological evolution is already opening the way for future high-speed optical networks, allowing the optimal tradeoff between spectral density and optical reach per deployed line system. Moreover the benefits of Flexgrid arrangement become readily evident when coupled with next-generation DSP, supporting for higher baud-rate transmission. Improving flexibility while increasing capacity of the photonics line allows giving service providers faster access to the latest advances in optical and silicon technologies on their switched optical http://dx.doi.org/10.1016/j.osn.2015.09.001 1573-4277/& 2015 Elsevier Ltd All rights reserved.
networks, and the potential evolution for the optical platforms it supports. From the beginning of 2013 many experimental trials have been proposed to demonstrate an evident technological advantage in high capacity optical transmission by different manufacturers, through the networks of the main International Global operators. Some of the more interesting results will be presented and discussed in the paper, especially concerning the linear properties of transmission.
2. Key innovation technology The push for 400G consolidates when 100G transport is an increasing industry success story. Hundreds of commercially deployed 100G systems across the globe and continued bandwidth growth pressures motivate the
156
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
industry increased focus on 400G and 1 Tb next generation of higher transmission systems to drive down cost of transported bit. Steady advance in optical and electronic transceiver technology supporting advanced modulation and spectral shaping techniques motivate the rate increasing. 400G data rate is a natural next step gives the current evolution of datacom and telecom transport interface speeds and implementation complexity. 400G and beyond R&D activities occurring now in order to build the technology innovations and sustainable ecosystem for a commercial future. Various ways to realize 400G transmission rate have been proposed:
increasing the symbol rate, increasing the number of bonded channels, increasing the modulation levels, or their combination For long haul distances, bonding a number of established PM-QPSK channels via WDM techniques can result in high spectrum efficiency and lower bandwidth requirements. For metro distances, higher‐level QAM formats can be used to achieve spectral efficiency that is greater than that of PM‐QPSK. Dual‐Carrier PM‐16QAM coding on 75– 100 GHz grid is an attractive solution to realize 400G line rates while at the same time relaxing the high speed bandwidth requirements of the electronic circuits and optoelectronic components. The coherent 200G/16QAM optical solution can provide many customer benefits: avoid saturating photonic lines and costly overlays, propagate with 10G NRZ, 40G DP-BPSK and 100G DP-QPSK, propagate over other vendor’s photonic line ( Fig. 1). Compared to today’s 100G chips, the new 400G Photonic Service Engine (PSE) designed by Alcatel-Lucent and tested in the described trial links delivers: 4 times the line rate; more than 2.6 times the spectral efficiency; more than 4 times the density; 33% lower power consumption and 25% better tolerance to fiber impairments. As a result, the device reaches the current technological limits of the market’s silicon for integrated digital signal processing in high-speed telecom applications ( Fig. 2). According to the key enable technologies to reach the high-speed fiber optics performances
DWDM solution to reach multiterabit/s capacity, with a
super-channel configuration in an optical multi carriers arrangement Reconfigurable Nodes and Flexible-grid operation Modulation format, advanced DSP, soft-decoding FEC and Coherent detection,
the main functionalities achieved during all the field-trial activities have been focused on:
High-speed 400G transmission over two aggregated sub-carriers 16QAM modulation scheme.
Integrated wave-shaping, pre-compensation and enhanced
optical signal to noise ratio (OSNR) performance, enabling very high tolerance to fiber non-linearities. Advanced algorithms enabling SW-programmable adaptive modulation format and bit-rate reconfiguration on the same optical channel including PDM-BSPK, PDM-QPSK and PDM-16QAM to optimize reach and capacity. Spectral efficiency improvement, demonstrating the benefits of “Flexible-grid Optical Networks” combining the best relation between distance and speed, according to the ITU-T Rec. G.694-1 [1]. Support for 50 GHz spacing at 400G for full compatibility with reconfigurable optical add-drop multiplexer (ROADM) networks.
With the optical technology available today, service providers must choose a specific modulation format for each signal speed. The 400G PSE forms the basis for software-defined flexible bit-rate and adaptive modulation line interfaces. A transponder based on this ASIC generation can generate a number of spectrally engineered modulation formats. This gives service providers the flexibility to choose the optimal modulation format for their distance and spectral efficiency requirements.
3. Capacity and bandwidth efficiency Bandwidth efficiency and increased role of digital signal processing in the optical layer have been the key elements to allow the multi-terabit transmission over the traditional channel planning and the flex-grid arrangement, recognized as the revolutionary technique for high capacity future systems. Considering, at the beginning, the limit for a linear channel with an additional Gaussian white noise distribution, the maximum capacity Cmax (maximum spectral efficiency) can be expressed as a function of channel width CW by the relation: C max ¼ CW log2 ð1 þSNRÞ In Fig. 3 the theoretical trend of the channel capacity as a function of OSNR OSNR ¼ CW=12:5 SNR is described, for different channel widths equal to the DWDM grid channel spacing. The theoretical OSNR degradation can be evaluated for a channel width of 37.5 GHz respect to the fixed 50 GHz WDM grid: 1 dB for a channel capacity of 100G, 2.5 dB for a capacity of 200G, 7 dB for a capacity of 400G. Software re-configurable QPSK/16QAM transponders guarantee excellent optical performances for both the adaptive rate (100 Gb/s and 200 Gb/s) in terms of OSNR and Q-factor. Figs. 4 and 5 show some significant measurements performed with standard production transponders. At the relative SD-FEC thresholds [BER ¼2.1 10 2], the deviations of measured OSNR respect to the theoretical
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
157
Fig. 4. Pre-FEC BER vs. OSNR measurements.
Fig. 1. QPSK and 16-QAM constellation diagrams.
Fig. 5. Q-factor vs. OSNR measurements.
Fig. 2. Evolve to 100G and beyond.
Fig. 3. Channel capacity vs. OSNR.
curves are really close: OSNRmeasured OSNRtheoretical ¼ 12:5 dB 11:0 dB ¼ 1:5 dB for 100G QPSK 20:5 dB 17:5 dB ¼ 3:0 dB for 200G16QAM The physics tradeoff of using 16QAM modulation vs. QPSK is that maximum reach decreases as bits/symbol
increase. For the 16QAM demonstrations, distances achieved are in the regional range of 400–600 km. Therefore, significant advances in FEC technology would be required in order to extend the 16QAM-based systems into long-haul and ultra-long-haul applications. Of course, the same OSNR deviations can be evaluated for the Q-factor limit of 6 dB, corresponding to the relative pre-FEC BER threshold of 2.1 10 2 for both the systems. The experimental links at high speed/capacity (41 Tb/s) have been demonstrated adjusting a super-channel of 7 200G carriers 16QAM modulated, with an optimized digital pulse shaping and operating with a spectral efficiency up to 5.7 bits/s/Hz. The 1.4 Tb/s super-channel is set up with seven 200G sub-channels generated from the software re-configurable transponders. In comparison with standard 50 GHz spacing, Fig. 6 shows the super-channel spectrum with sub-channel spacing at 50 GHz and 35 GHz respectively. A raised-cosine filter has been used for TX pulse shaping, reducing the crosstalk and the inter-symbol interference [2]. The spectral reduction from 350 GHz to 245 GHz represents an effective bandwidth occupancy improvement of about 30% (Fig. 7). In the described field trial, the 1.4 Tb/s Alien Superchannel propagates alongside the existing 40G and 100G native wavelengths over the trial link. Fig. 8 shows the measured optical spectrum of the native wavelengths and 1.4 Tb/s Alien Superchannel after 410 km transmission. The
158
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 10. C-band capacities in split spectrum and gridless arrangement.
Fig. 6. 1.4 Tb/s super-channel spectrum.
Fig. 7. Spectral reduction in flexgrid arrangement.
Fig. 8. Optical wavelengths.
spectrum
of
the
1.4 Tb/s
and
native
40G/100G
figure is indicative only and directly downloaded from an optical spectrum analyzer, so the bad quality depends from the instrument imaging processing. Fig. 9 shows the pre-FEC BER performance of the Alien Superchannel over 410 km for sub-channel spacing at 50 GHz, 40 GHz, 37.5 GHz, 36 GHz and 35 GHz respectively. All the seven 200G sub-channels achieved pre-FEC BER of better than 2 10 2, which gives error free performance after FEC. At 50 GHz spacing, the Alien Superchannel has the lowest pre-FEC BER as expected. As the channel spacing narrows, the pre-FEC BER degrades due to the increasing crosstalk between the adjacent channels. Higher order modulation will certainly be a useful tool to allow service providers to optimize the total spectral efficiency for certain routes. But the penalty for this is reduced reach. Unfortunately, higher order modulation on its own is not a solution in moving to Terabit capacities, and there is no single correct modulation technique for any given route. Once again, this highlights the need for a multi-carrier super-channel implementation with adaptive coherent modulation so that service providers can optimize the combination of reach and spectral efficiency without having to order multiple part numbers from their system supplier [3]. Fig. 10 illustrates the total transmitted capacity on the whole C-band for a selection of modulation schemes, in a fixed split spectrum condition and in a gridless arrangement. Again, for all the considered modulation schemes an evident advantage of 33.33% is demonstrated in a gridless configuration with respect to the standard fixed split spectrum arrangement at n 50 GHz. Fig. 11 shows the normalized reach (assuming the QPSK application as standard reference) for the different schemes: modulation techniques likes 16QAM may be limited to regional network use. This comparison is illustrative only: not all modulation types shown are practical, and other modulation types may be available in final products.
4. Flexgrid arrangement
Fig. 9. Performance of the Alien Superchannel over 410 km.
In the traditional approach to WDM, the optical transmission spectrum is carved into a number of nonoverlapping wavelength (or frequency) bands with spectrum spacing of 50 GHz (or 100 GHz), as specified by ITU-T standards, and each wavelength supports a single communication channel operating at a certain fixed rate
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 11. Normalized reach for different modulation formats.
(e.g., 10 Gb/s, 40 Gb/s, or 100 Gb/s). This is referred to as the fixed grid. Recent trends in optical transmission technologies such as coherent detection and digital signal processing are opening new horizons for novel flexible transmission systems and optical devices such as bandwidth-variable Optical Cross-connects or bandwidthvariable ROADMs. A more practical spectrum management for optical networks is the so-called flexible grid, in which the spectrum grid granularity is much finer than the ITU-T grid and a lightpath’s spectrum can span across several frequency slices. [4] Key technology elements to achieve flexible line are:
Flexgrid, utilizing reduced spacing in order to increase spectral efficiency and introduce new level of flexibility
Flexible wave-shaping, enabling a further increment in network capacity
Coherent technology, pick and choose modulation depending on networking need
Programmable capacity Resiliency and control
Comparison of fixed WDM and flexible grid approaches has been evaluated in terms of spectral efficiency improvement and spectral distribution optimization for future “elastic” or rate-variable optical networks. Fig. 12 compares the spectral efficiency for the different 200G sub-carriers spacing and the corresponding spectral efficiency improvement percentage respect to the fixed 50 GHz channel spacing solution. With the 35 GHz channel spacing limit, the spectral efficiency grows up to 6 bit/s/Hz with an improvement of more than 40%. A further comparison of fixed WDM and flexible grid approaches can be evaluated in terms of spectrum occupation and transmitted capacity. In the fixed WDM network, the total available spectrum of 4.4 THz is divided in 88 slices of 50 GHz each. In the flexible grid arrangement we consider a grid granularity of 6.25 GHz or 12.5 GHz, which leads to 704 or 352 frequency slices. The effective signal baud-rate of 28 Gbaud (symbol rate¼32.5 Gbaud and roll-off¼0.16) matches with both 31.25 GHz or 37.5 GHz spacing, corresponding to 5 slices of 6.25 GHz width or 3 slices of 12.5 GHz width, respectively. In both the fixed grid and flexible grid arrangement, we consider adaptive digital coherent transceivers supporting
159
polarization multiplexed BPSK, QPSK and n-QAM modulation formats, with n ¼8, 16, 32, 64, 128. The transceiver utilizes the most efficient modulation format compatible with the distance to be covered by the lightpath, with the associated bit-rate of each single traffic request. Introducing a flexgrid arrangement allows a significant improvement in terms of spectral efficiency, especially using a channel bandwidth of 5 6.25 GHz (as demonstrated in Fig. 13) where, for a bit-rate of 200 Gb/s, a total efficiency of 6.4 bit/s/Hz is reached with a 16QAM modulation format, comparing with 4 bit/s/Hz using a 50 GHz WDM grid. A possible future implementation of 400G singlecarrier with an available 128QAM scheme in a channel bandwidth of 5 6.25 GHz per carrier could reach a spectral efficiency of about 13 bit/s/Hz, instead of 8 bit/s/ Hz using a 50 GHz grid. The spectrum savings (or spectral efficiency increasing) in flexible grid scenario with respect to the fixed WDM grid are uniquely motivated to reducing the frequency slice width. The slight reductions in the overall spectrum occupied are obtained due to finer capacity granularity, which allows for better adaptation of the number of transceivers to the traffic volume to be served, thus reducing the amount of unused capacity on the installed devices and consequently leading to more efficient spectrum utilization [5,6]. The same results can be seen in terms of total transmitted capacity in the optical bandwidth. Current fiber usage is limited by the fixed 50 GHz ITU-T grid (Fig. 14): 88 100G channels ðPDM QPSK
50GHzÞ-8:8 Tbps
44 400G channels ð2 PDM 16QAM50GHzÞ -17:6 Tbps Flex grid brings improved spectrum usage. Actual grid can be selected to match the spectrum of the signal, with frequency slots of nx12.5 GHz or n 6.26 GHz. Superchannels wider than 50 GHz are feasible: 116 100Gsuper chs:ðPDM QPSK
37:5 GHzÞ
-11:6Tbps 140 100Gsuper chs:ðPDM QPSK
31:25 GHzÞ
-14:0Tbps 58 400Gsuper chs:ð2 PDM 16QAM
75 GHzÞ
-23:2Tbps 70 400Gsuper chs:ð2 PDM 16QAM6
2:5 GHzÞ
-28:0Tbps The total transmitted capacity (described in Fig. 15 as a function of different bit-rate) has been evaluated over the complete C-band frequency window (4400 GHz, equivalent to 88chs at 50 GHz). The bit rate Br is a function of the number of bits/ symbol M (modulation factor) given by the modulation scheme, the symbol rate Sr and the number of (sub-) carriers SC: Br ¼ M Sr SC With higher level modulation formats, the number of bits per symbol is increased. E.g. for DP-m-QAM
160
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 12. Spectral efficiency comparison.
Fig. 13. Spectral efficiency comparison.
Fig. 14. Example of fixed and flexible grid channels distribution.
Fig. 16. Transmitted capacity using modulation schemes optimization.
multiple of the spectrum slice width (n ¼5BW¼31.25 GHz) better fit the network configurations required to achieve optimal spectrum assignment. In Fig. 16 we can see the total transmitted capacity for different client bit-rate optimized with the suitable DP-mQAM modulation format. Considering a bit-rate of 200 Gb/s per carrier (6.25 GHz spaced), a maximum capacity of 28 Tb/s could be transmitted in the total available bandwidth. We will first look at the gains from using Flexgrid when compared to inverse multiplexing. Inverse multiplexing is one technique to enable higher bitrate services to continue to use a fixed grid. In this approach, multiple lower bitrate channels are multiplexed together to form a higher aggregated bitrate. For example, a 400 Gbit/s signal could be transmitted using 4 100 Gbit/s sub-signals each of which could continue to fit within a 50 GHz channel. This approach does have the drawback that more spectrum is utilized than is absolutely necessary: using 4 100 Gbit/s channels uses 200 GHz of optical spectrum compared to just 75 GHz for transferring a native 400 Gbit/s signal. The efficiency improvement percentage for flexible spectrum over a point-to-point link is shown in Fig. 17, assuming a 50 GHz grid for fixed DWDM and super-channels with different Spectral Width (SW) for the applied modulation schemes. The efficiency gain of using native transponders compared to inverse multiplexing is easy to calculate numerically for a point-to-point link using the formula given in the following equation: Efficiency
Fig. 15. Total transmitted capacity in fixed and flexgrid arrangement.
modulation schemes, the modulation factor is given as M ¼ 2 log2 ðmÞ A finer grid granularity (nx6.25 GHz) combined with transceivers working at baud rates close to an integer
gain ¼
SW f ixed ðSW f lexgrid þ SW guardband Þ SW f lexgrid þ SW guardband
SW stands for the spectral width of the transponder in Flexgrid and Fixed grid scenarios. For the Flexgrid scenario, a small amount of spectrum (76.25 GHz) needs to be placed between signals to form a guard band. Of course, the super-channel implementation with a multi-carrier arrangement exhibits the best efficiency gain. The 400G solution (4 100G) or 1Tb (10 100G) operating in an optimized flexgrid bandwidth arrangement of n 12.5 GHz would reach the higher efficiency gain (up to 150%) respect to the fixed n 50 GHz spectral width. Parameters to increase the capacity are mainly the number of carriers, the modulation format and the baudrate. The represented curves in Fig. 18 justify and demonstrate the bit rate improvement per single carrier with the usage of higher baud-rate solution and the
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 17. Flex-grid efficiency for different rate and modulation formats.
Fig. 18. How to increase capacity.
Fig. 19. Normalized Capacity*Reach product.
high spectrum efficiency with higher order modulation schemes. The Capacity*Reach product (normalized at the achieved value with a QPSK modulation scheme) as a function of the total transmitted C-Band capacity is described in Fig. 19. The transmission of 80 200G carriers, 16QAM modulated, allows a total capacity of 16 Tb/s in the C-band, with a maximum reach equivalent to 40% of the nominal reach achieved with the 100G QPSK modulated carriers, performing the half of the total transmitted capacity. The required OSNR depends from the adopted modulation scheme and the system symbol rate. An example of
161
Fig. 20. Required OSNR for different modulation schemes.
Fig. 21. Spectral efficiency and total capacity for different 400G arrangements.
OSNR distribution for working baud-rate at 28 Gbaud and 56 Gbaud is calculated in Fig. 20 for some amplitude modulation schemes. The OSNR degradation between the relative symbol rate at the same modulation format is 3 dB. Today, available technologies beyond 100G using higher order constellation than QPSK have led to reach limitations, mainly as regards the maximum transmission reach. PM-16QAM technology 200G or 400G superchannel (2 200G sub-carrier) is often proposed by vendors, despite of 7 dB extra-OSNR required vs. 100G PM-QPSK, more sensitive to filtering caused by cascade of ROADMs, a limited 400 km maximum reach instead of a significant spectral efficiency improvement (from 2 to 5.33 bit/s/Hz). In Fig. 21 the spectral efficiency and the total transmitted capacity on the whole C-band have been calculated for different 400G multi-carriers spectral configurations, assuming a spectral bandwidth optimized with n 12.5 GHz slices for each modulation scheme. The implemented 2 200G (16QAM modulation scheme) in 75 GHz channel spacing performs the higher spectral efficiency (5.33 bit/s/Hz) and the maximum transmitted capacity (23.2 Tb/s) corresponding to 58 equivalent 400G channels in the total optical available bandwidth. In Figs. 22 and 23 the constellation diagrams for the DP-16QAM modulation scheme (both polarizations) are shown. Transmit wave-shaping compresses signals approaching the Nyquist limit. The optimization of the baud-rate and the filter width allows achieving better performances:
162
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
Fig. 22. 16QAM constellation (in back-to-back configuration).
Fig. 23. 16QAM constellation (after transmission over 400 km G.652 fiber).
compressed signals on 37.5 GHz grid see similar performance as standard signals on 50 GHz grid (see Fig. 24). In Fig. 25 the transfer function of the WSS filter with a bandwidth of 75 GHz has been optimized for the transmission of a flexible super-channel of 2 200G carriers, 37.5 GHz spaced. The image has been downloaded directly from the optical spectrum analyzer.
5. OSNR penalties vs. sub-channel spacing To achieve optimized performance, the spectral shaping filter is optimized corresponding to different channel spacing. A digital raised-cosine pulse shaping has been implemented with a roll-off optimized for a minimum BER floor for each carrier spacing value. Computation of OSNR penalties has been evaluated for sub-channels spacing down to 35 GHz. Fig. 26 shows the OSNR penalty curve for a sub-channel, which is at the middle of the Alien SuperChannel and suffers most from both linear and nonlinear
crosstalk from its adjacent channels. Compared to 50 GHz spacing, the OSNR penalty is within 0.5 dB for sub-channel spacing down to 37.5 GHz. Beyond this the optical penalty increases steeply. At 35 GHz spacing, error free performance is achieved with no margin. For channel spacing at 37.5 GHz or wider we measured stable, long term error free performance. The raised-cosine pulse shaping and the roll-off have been optimized for a minimum BER floor for each carrier spacing value, as indicated in Fig. 27 where are shown the correspondent pre-FEC BER curves. The transmit (TX) DSP wave-shaper can shape the transmit spectrum into any given shape based on filter configuration. This capability allows service providers to make trade-offs between spectral efficiency and non-linearity tolerance. It can optimize the overall spectral content of the signal, offering a higher degree of tolerance to noise and minimize crosstalk effects [7]. The bandwidth reduction of the optical signal allows denser channel spacing and increases capacity when
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
163
37.5 GHZ 50 GHZ
Fig. 24. Spectral distribution for different signal shaping.
Fig. 27. Pre-FEC BER vs. OSNR for different carrier spacings. Fig. 25. WSS filter (BW ¼75 GHz).
Fig. 28. OSNR penalty vs. normalized channel spacing. Fig. 26. OSNR penalty for different channel spacings.
flexible grids are employed, introducing OSNR penalties. Fig. 28 shows a plot of sensitivity as a function of the normalized sub-channel spacing Δf/Sr when using an
optimized TX digital filter. When the sub-channel spacing is less than 1.1 times the symbol rate the performances become unacceptable in terms of OSNR penalties. In order to match the granularity of the flexible grid, the channel bandwidth BW must be an integer multiple n
164
C. Colombo / Optical Switching and Networking 19 (2016) 155–164
6. Conclusions
Fig. 29. Transmission reaches for different modulations and channel spacing.
of the slice width. The choice of n impacts on the performance of the transceiver: the higher the Sr/BW ratio, the higher the efficiency of the spectrum utilization, but the lower the maximum transmission reach length that can be covered without need of electronic regeneration. Assuming a symbol rate Sr ¼28 Gbaud and considering the channel bandwidths BW1 ¼50 GHz, BW2 ¼37.5 GHz (3 12.5 GHz) and BW3 ¼31.25 GHz (5 6.25 GHz): Sr =BW 1 ¼ 0:56;
Sr =BW 2 ¼ 0:75;
Sr =BW 3 ¼ 0:9
Fig. 29 reports the maximum reaches for different modulation formats (at the same symbol rate of 28 Gbaud) and the considered ratios. For a 16-QAM scheme the relative distance reduction is little more than 15% for a BW¼37.5 GHz and about 20% for a BW¼31.25 GHz, respect to the conventional WDM grid of 50 GHz. The reduction in reach is mainly caused by additional crosstalk due to adjacent channel coherent interference, and is more pronounced with lower modulation formats. The optical pre-filtering increases the system spectral efficiency, allowing the allocation of a higher number of channels in the same bandwidth and maintaining the modulation format at a penalty cost of transmission performance due to the intersymbol interference caused by the filtering. Thus, pre-filtering systems should be designed considering a trade-off between spectral efficiency, performance and maximum reach. This is an important factor when we consider that the future of optical networks tends to flexible applications of bit rate, modulation formats and channel bandwidth. [8]
The next generation of Dense Wavelength Division Multiplexing (DWDM) networks are likely to use Flexgrid arrangement, providing operators with additional flexibility when assigning spectrum compared to traditional DWDM networks using the 50 GHz ITU grid. This additional flexibility will allow faster transponders that utilize high spectral efficiency modulation techniques, like multilevel m-QAM schemes. From the use of these new spectrum efficient modulation formats and finer control over spectrum allocations, a key benefit that Flexgrid offers to the network operators is that their DWDM networks can carry more traffic with optimized spectral efficiency. High speed technology became essential for realizing greater network capacity and enabling network operators to meet the increasing bandwidth challenges from new generation superfast devices, services and applications. Coherent technologies beyond 100G are now focusing on higher level modulation formats and multiple sub-carriers/channels, using super-channels to achieve Terabit transmission. Within the next years, 400G is set to become the dominant backbone technology in terms of capacity shipped, replacing longtime incumbent 10G and 40G optical lines. OTN switching, ASIC chip design, coherent detection and soft-decision FEC are among the best innovations that will make 400G a success. The demonstration of these implemented features was evident in the awareness of the high technology level inherent to the proposed solution.
References [1] ITU-T Recommendation G.694-1, Spectral grids for WDM applications: DWDM frequency grid. [2] Yu Rong Zhou, et al., 1.4 Tb Real-Time Alien Superchannel Transport Demonstration over 410 km Installed Fiber Link Using Software Reconfigurable DP-16QAM/QPSK, pdp OFC, 2014. [3] T. Xia, et al., High-capacity optical transport networks, IEEE Commun. Mag. 50 (11) (2012) 170–178. [4] O. Gerstel, et al., Elastic optical networking: a new dawn for the optical layer? IEEE Commun. Mag. 50 (2) (2012) s12–s20. [5] C. Rottondi, et al., Optical ring metro networks with flexible grid and distance-adaptive optical coherent transceivers, Bell Labs Tech. J. 18 (3) (2013) 95–110. [6] C. Rottondi, et al., Routing, modulation level, and spectrum assignment in optical metro ring networks using elastic transceivers, J. Opt. Commun. Netw. 5 (4) (2013) 305–315. [7] P. Winzer, et al., Penalties from In-Band Crosstalk for Advanced Optical Modulation Formats, in: Proceedings of the ECOC, 2011. [8] G. Bosco, et al. Investigation on the robustness of a Nyquist-WDM terabit superchannel to transmitter and receiver non-idealities, in: Proceedings of the ECOC, 2010.