140 journal of lightwave technology, vol. … · t. rahman is with the coriant r&d gmbh, munich...

11
140 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 33, NO. 1, JANUARY 1, 2015 Digital Pre-Emphasis in Optical Communication Systems: On the Nonlinear Performance Danish Rafique, Talha Rahman, Antonio Napoli, and Bernhard Spinnler Abstract—Digital signal pre-processing at the transmitter is a key enabler for next-generation optical transceivers. One of the major challenges faced by these transponders is the limited res- olution and 3 dB electrical bandwidth of the digital-to-analog converters (DAC), severely limiting the transmission performance. On the other hand, these spectrally efficient flexible transport net- works are extremely sensitive to nonlinear channel impairments, essentially limiting the maximum system reach. In this paper, we employ a simple digital pre-emphasis (DPE) algorithm to miti- gate DAC-induced signal distortions, and further report on the im- pact of DPE algorithm on nonlinear transmission performance. We demonstrate, both numerically and experimentally, that DPE not only counteracts DAC low-pass response, allowing for higher baud rate transmission, but also enables better nonlinear channel toler- ance. In particular, we first establish the maximum transmittable baud rates in back-to-back configuration for polarization multi- plexed m-state quadrature amplitude modulation (PM-mQAM) formats, allowing up to 50, 46 and 44 Gbaud transmission for PM-4QAM, PM-8QAM, and PM-16QAM, respectively assuming typical DAC specifications of 16 GHz 3 dB bandwidth and 5.5 ef- fective number of bits. Furthermore, we show that the performance improvements from DPE increase with increasing modulation or- der and decreasing span count, with maximum improvements of up to 2.2 dB. In a conventional standard single mode fiber trans- mission link with lumped amplification, employing DPE, we re- port maximum reach of up to 10 000, 5000, and 3000 km, for PM-4QAM, PM-8QAM, and PM-16QAM, respectively. More- over, at pre-FEC bit-error-rate, relative reach improvements from DPE are found to be up to 21%, 25%, and 27%, respectively. Fi- nally, we show that DPE gains are constant beyond a minimum required channel spacing, with minimum spacing requirements of 1.1 × baud rate, 1.15 × baud rate, and 1.2 × baud rate for PM-4QAMPM-16QAM, respectively. Index Terms—Coherent detection, digital-to-analog converter, digital signal processing, Kerr effect, pre-emphasis, Nyquist. I. INTRODUCTION T ODAY’s commercial long-haul optical products employ 100 Gb/s polarization multiplexed quadrature phase mod- ulation (PM-QPSK) transponders with powerful receiver side digital signal processing (DSP) algorithms, fully capable to Manuscript received October 20, 2014; revised November 12, 2014 and De- cember 2, 2014; accepted December 2, 2014. Date of publication December 3, 2014; date of current version January 23, 2015. This work was supported by the German Federal Ministry of Education and Research (BMBF) under Grant 01BP12300A, EUREKA-Project SASER, and by the FP-7 project, IDEALIST, under Grant 317999. D. Rafique, A. Napoli, and B. Spinnler are with the Coriant R&D GmbH, Mu- nich 81541, Germany (e-mail: danish.rafi[email protected]; antonio.napoli@ coriant.com; [email protected]). T. Rahman is with the Coriant R&D GmbH, Munich 81541, Germany, and also with COBRA Research Institute, Eindhoven University of Technology, 5612 AZ Eindhoven, The Netherlands (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JLT.2014.2378374 mitigate linear channel response and phase noise effects [1], [2]. The next-generation of optical transmission products not only aims to employ advanced spectrally efficient modula- tion formats (e.g., polarization multiplexed m-state quadrature amplitude modulation, PM-16QAM), but also improve on the flexibility of network architecture [3]–[6]. Flex-grid or software- defined network architectures have thus gained substantial in- terest, employing a switchable network configurator, allowing to adapt to the appropriate modulation format, bit-rate, channel grid, etc., for a given link configuration [7]–[10]. In this context, digital-to-analog converters (DAC) have be- come a key constituent of cutting edge transmission systems [11], as they allow for features like adaptive waveform gen- eration, digital pulse shaping, digital pre-compensation of lin- ear and nonlinear response, etc. [12]–[14]. Current generation DACs typically have nominal resolution of 8 bits—effective number of bits (ENoB) <5.5, specified 3 dB bandwidth of less than 16 GHz, and sampling rates in the order of 65 GS/s [15]–[17]. While sampling rates are adequate for negligible per- formance penalties (1.2 × baud rate with digital pulse shaping) [18], the low pass response and limited DAC resolution induce severe inter-symbol interference [19]–[20]. This is the key reason why maximum baud rates with inte- grated commercial DACs are still limited to around 30 Gbaud [2], [21], [22]. In order to enable higher baud rate transmission systems, allowing for higher forward error correction (FEC) overhead and/or saving on transponder count, DAC limita- tions need to be addressed. Consequently, digital pre-emphasis (DPE), a well-known concept in radio frequency communica- tions [23], [24], has recently been used in DAC based coherent optical systems, extending the maximum baud rates well beyond typical channel rates [25]–[30] in conventional wavelength di- vision multiplexed (WDM) system architecture. On the other hand, channel nonlinearities are known to limit the maximum potential transmission performance, even more critically for multi-level advanced modulation formats [31], [32]. Recently, various techniques have been proposed to mini- mize such effects, e.g., digital back-propagation [33], [34], spec- tral inversion [35], [36], perturbation based nonlinear mitigation [37], [38], etc., however, either at the cost of additional hardware complexity [39]–[41] or modification of network architecture and control plane. In this paper, we present the second part of our recently pub- lished work on the DAC requirements for Terabit applications [42], and investigate the impact of DPE on nonlinear channel performance, both numerically and experimentally. In particu- lar, we establish maximum potential baud rates for PM-mQAM formats, and report on the maximum transparent reach, with and without DPE, at a given pre-FEC threshold. Furthermore, we 0733-8724 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

Upload: nguyenkhuong

Post on 29-Aug-2018

213 views

Category:

Documents


0 download

TRANSCRIPT

140 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 33, NO. 1, JANUARY 1, 2015

Digital Pre-Emphasis in Optical CommunicationSystems: On the Nonlinear Performance

Danish Rafique, Talha Rahman, Antonio Napoli, and Bernhard Spinnler

Abstract—Digital signal pre-processing at the transmitter is akey enabler for next-generation optical transceivers. One of themajor challenges faced by these transponders is the limited res-olution and −3 dB electrical bandwidth of the digital-to-analogconverters (DAC), severely limiting the transmission performance.On the other hand, these spectrally efficient flexible transport net-works are extremely sensitive to nonlinear channel impairments,essentially limiting the maximum system reach. In this paper, weemploy a simple digital pre-emphasis (DPE) algorithm to miti-gate DAC-induced signal distortions, and further report on the im-pact of DPE algorithm on nonlinear transmission performance. Wedemonstrate, both numerically and experimentally, that DPE notonly counteracts DAC low-pass response, allowing for higher baudrate transmission, but also enables better nonlinear channel toler-ance. In particular, we first establish the maximum transmittablebaud rates in back-to-back configuration for polarization multi-plexed m-state quadrature amplitude modulation (PM-mQAM)formats, allowing up to 50, 46 and 44 Gbaud transmission forPM-4QAM, PM-8QAM, and PM-16QAM, respectively assumingtypical DAC specifications of 16 GHz −3 dB bandwidth and 5.5 ef-fective number of bits. Furthermore, we show that the performanceimprovements from DPE increase with increasing modulation or-der and decreasing span count, with maximum improvements ofup to ∼2.2 dB. In a conventional standard single mode fiber trans-mission link with lumped amplification, employing DPE, we re-port maximum reach of up to ∼10 000, ∼5000, and ∼3000 km,for PM-4QAM, PM-8QAM, and PM-16QAM, respectively. More-over, at pre-FEC bit-error-rate, relative reach improvements fromDPE are found to be up to 21%, 25%, and 27%, respectively. Fi-nally, we show that DPE gains are constant beyond a minimumrequired channel spacing, with minimum spacing requirementsof 1.1 × baud rate, 1.15 × baud rate, and 1.2 × baud rate forPM-4QAM→PM-16QAM, respectively.

Index Terms—Coherent detection, digital-to-analog converter,digital signal processing, Kerr effect, pre-emphasis, Nyquist.

I. INTRODUCTION

TODAY’s commercial long-haul optical products employ100 Gb/s polarization multiplexed quadrature phase mod-

ulation (PM-QPSK) transponders with powerful receiver sidedigital signal processing (DSP) algorithms, fully capable to

Manuscript received October 20, 2014; revised November 12, 2014 and De-cember 2, 2014; accepted December 2, 2014. Date of publication December 3,2014; date of current version January 23, 2015. This work was supported bythe German Federal Ministry of Education and Research (BMBF) under Grant01BP12300A, EUREKA-Project SASER, and by the FP-7 project, IDEALIST,under Grant 317999.

D. Rafique, A. Napoli, and B. Spinnler are with the Coriant R&D GmbH, Mu-nich 81541, Germany (e-mail: [email protected]; [email protected]; [email protected]).

T. Rahman is with the Coriant R&D GmbH, Munich 81541, Germany, andalso with COBRA Research Institute, Eindhoven University of Technology,5612 AZ Eindhoven, The Netherlands (e-mail: [email protected]).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/JLT.2014.2378374

mitigate linear channel response and phase noise effects [1],[2]. The next-generation of optical transmission products notonly aims to employ advanced spectrally efficient modula-tion formats (e.g., polarization multiplexed m-state quadratureamplitude modulation, PM-16QAM), but also improve on theflexibility of network architecture [3]–[6]. Flex-grid or software-defined network architectures have thus gained substantial in-terest, employing a switchable network configurator, allowingto adapt to the appropriate modulation format, bit-rate, channelgrid, etc., for a given link configuration [7]–[10].

In this context, digital-to-analog converters (DAC) have be-come a key constituent of cutting edge transmission systems[11], as they allow for features like adaptive waveform gen-eration, digital pulse shaping, digital pre-compensation of lin-ear and nonlinear response, etc. [12]–[14]. Current generationDACs typically have nominal resolution of 8 bits—effectivenumber of bits (ENoB) <5.5, specified −3 dB bandwidth ofless than 16 GHz, and sampling rates in the order of 65 GS/s[15]–[17]. While sampling rates are adequate for negligible per-formance penalties (∼1.2× baud rate with digital pulse shaping)[18], the low pass response and limited DAC resolution inducesevere inter-symbol interference [19]–[20].

This is the key reason why maximum baud rates with inte-grated commercial DACs are still limited to around 30 Gbaud[2], [21], [22]. In order to enable higher baud rate transmissionsystems, allowing for higher forward error correction (FEC)overhead and/or saving on transponder count, DAC limita-tions need to be addressed. Consequently, digital pre-emphasis(DPE), a well-known concept in radio frequency communica-tions [23], [24], has recently been used in DAC based coherentoptical systems, extending the maximum baud rates well beyondtypical channel rates [25]–[30] in conventional wavelength di-vision multiplexed (WDM) system architecture.

On the other hand, channel nonlinearities are known to limitthe maximum potential transmission performance, even morecritically for multi-level advanced modulation formats [31],[32]. Recently, various techniques have been proposed to mini-mize such effects, e.g., digital back-propagation [33], [34], spec-tral inversion [35], [36], perturbation based nonlinear mitigation[37], [38], etc., however, either at the cost of additional hardwarecomplexity [39]–[41] or modification of network architectureand control plane.

In this paper, we present the second part of our recently pub-lished work on the DAC requirements for Terabit applications[42], and investigate the impact of DPE on nonlinear channelperformance, both numerically and experimentally. In particu-lar, we establish maximum potential baud rates for PM-mQAMformats, and report on the maximum transparent reach, with andwithout DPE, at a given pre-FEC threshold. Furthermore, we

0733-8724 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

RAFIQUE et al.: DIGITAL PRE-EMPHASIS IN OPTICAL COMMUNICATION SYSTEMS: ON THE NONLINEAR PERFORMANCE 141

also establish minimum channel spacing requirements for highbaud rate transmission systems, and show that DPE gains areindependent of channel spacing beyond the minimum requiredchannel grid. In Section II, the principle of digital pre-emphasis,and the proposed algorithm are briefly explained. In Section III,the numerical transmission setup is described with various pa-rameter settings and system configurations. Section IV reportson maximum transportable baud rates, and presents numericalresults, establishing DPE gains as a function of transmissionreach for various modulation formats. In addition, performancetradeoffs between baud rate induced performance degradationand maximum attainable FEC are discussed. Furthermore, weestablish channel spacing requirements, both with and withoutDPE. In Section V, the experimental setup is explained, andSection VI discusses experimental confirmation of numericalobservations, for a 36 Gbaud PM-16QAM 1.15 Tb/s WDMtransmission system. Finally, Section VII draws conclusions.

II. PRINCIPLE OF DIGITAL PRE-EMPHASIS

In this section we briefly introduce the digital pre-emphasisalgorithm to mitigate the low pass response of DAC, taking intoaccount the finite resolution of this device. A signal, S(f), atthe transmitter was passed through a pre-emphasis stage, beforeinflowing to the DAC, with output signal D(f). The pre-emphasisfilter function P(f) was approximated using a modified minimummean square like criterion, given as follows:

P (f) =DDAC,DESIRED (f)DDAC,REAL (f) + no

(1)

where no is the noise adjustment term, DDAC,REAL (f) isthe real transfer function of DAC, and DDAC,DESIRED (f)is the desired DAC transfer function. In our simulations weimplemented the pre-emphasis filter in the frequency do-main using the overlap/save technique with an overlap size of128 taps keeping the focus mainly on performance and noton complexity. In practice, the filter structure may be cho-sen to balance increasing complexity and performance im-provements [18]. We modeled the realistic and desired DACresponse using third order Bessel and Gaussian functions, re-spectively. Note that DAC transfer function was chosen basedon commercial DAC measurements and available literature[43]. In case of DDAC,REAL (f) and DDAC,DESIRED (f), the−3 dB DAC bandwidth, BDAC , was set to BDAC,REAL andBDAC,DESIRED , respectively. Fig. 1 shows a typical optimiza-tion contour, between BDAC,DESIRED and no , for 44 GbaudPM-16QAM signal after 400 km transmission, with −3 dBDAC bandwidth of 16 GHz and ENoB of 5.5 bits. It can be seenthat the minimum required BDAC,DESIRED was found to be∼>23 GHz with subtle dependency on the noise optimizationcoefficient. The optimum desired bandwidth depends upon theDAC specification, the baud rate, and the modulation format.

The calculated pre-emphasis filter response, using Eq. (1),was tuned using feedback from receiver in back-to-back systemconfiguration. Note that this optimization only needs to be doneat system startup or during factory calibration. More details ofthe presented model can be found in [42].

Fig. 1. Pre-emphasis filter response optimization (Q-factor as a function ofjoint optimization of desired DAC bandwidth and noise adjustment coefficient).PM-16QAM (44 Gbaud) after 400 km, with real DAC parameters of 16 GHzbandwidth and 5.5 effective number of bits. Blue arrows show the minimumrequired optimum bandwidth.

Fig. 2. Simulation setup for back-to-back and link transmission of m-statePM-QAM using digital pre-emphasis. Various baud rates are generated for eachmodulation format (30 to 62 Gbaud). Also, different reach and channel spacingvalues are considered. (DAC specifications are assumed to be 16 GHz and 5.5ENoB).

It is worth mentioning that, in principle, DAC induced im-pairments may be compensated using a linear equalizer at thereceiver node. However, since the transmitted signal is impairedby channel noise as well, a receiver based post-emphasis typi-cally results in additional digitally enhanced noise, compared toonly quantization noise limitation in pre-emphasis stage. Fur-thermore, as discussed in this article, the channel nonlinearitiesmay not be suppressed by the DPE filter, if the equalizer isemployed at the receiver. Nonetheless, there have been recentreports on receiver based compensation of DAC effects, em-ploying a series of algorithms, typically based on maximumlikelihood probability theory [19], [44].

III. NUMERICAL TRANSMISSION CONFIGURATION

Fig. 2 shows the simulation setup considered in this work.PM-m-state QAM (m = 4, 8, 16) signals were generated at vari-able baud rates, where both polarization states were modulatedindependently using de-correlated bit sequences. We employed

142 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 33, NO. 1, JANUARY 1, 2015

WDM homogeneous transmission scenario, where the neigh-boring channels were considered to be similar to central test-channel, and the number of neighbors was always fixed to 12.In order to minimize the impact of linear sub-carrier crosstalk,and allow dense spectral grid, spectral shaping was applied indigital domain, where the root-raised cosine (RRC) filter roll-offcoefficient was fixed at 0.2. Moreover, the spacing within thesubcarriers was also optimized, and fixed at ∼1.2 × baud rate(unless mentioned otherwise). Each digital sequence was de-multiplexed separately into two output symbol streams whichwere symbol mapped, digital shaped, pre-emphasized using al-gorithm described in Section II, and processed by DAC to gen-erate drive signals for in-phase and a quadrature phase carriers.Note that results are presented both with and without digital pre-emphasis. DACs were modelled considering −3 dB bandwidthof 16 GHz, ENoB of 5.5 bits, and sampling rate of 80 GS/s. Notethat the DAC sampling rate dependency is not investigated inthis study, and the selected sampling rate assures negligible per-formance penalty for the highest baud rate considered (∼1.3 ×62 Gbaud) [18]. The optical transmitters consisted of continuouswave LASER sources, followed by two nested Mach-Zehndermodulator structures for x- and y-polarization states. The twopolarization states were combined using a polarization beamcombiner.

The signals were multiplexed, and propagated over stan-dard single mode fiber (SSMF), with loss coefficient of0.21 dB/km, dispersion coefficient of 16.8 ps/nm/km, disper-sion slope of 0.058 ps/nm2/km, polarization mode dispersionof 0.06 ps/sqrt(km), and fiber nonlinearity coefficient of 1.141/W/km. As shown in Fig. 2, the transmission link consistedof 80 km spans, no inline dispersion compensation and single-stage erbium doped fibre amplifiers (EDFAs). Each amplifierstage was modelled with a 5 dB noise figure and the total ampli-fication gain was set to be equal to the total loss in each span. Incase of back-to-back transmission configuration, additive whiteGaussian noise was added at the receiver, and the received op-tical signal-to-noise ratio (OSNR) was fixed at 15.2 dB, 19 dBand 22 dB for PM-4 QAM, PM-8 QAM, and PM-16 QAM,respectively, allowing theoretical bit-error-rate (BER) of ∼10−4

at lowest considered baud rate of 30 Gbaud.At the polarization diversity receiver, the central channel was

de-multiplexed using coherent channel selection, and detectedusing four balanced detectors to give the baseband electricalsignal. The signal was then digitized using analog-to-digital con-verter (ADC), with same settings as the DAC. Transmission im-pairments were digitally compensated using conventional DSPblocks, clock recovery, frequency domain dispersion compen-sation, polarization de-multiplexing, and carrier recovery [45].Finally, the symbol decisions were made, and the performanceassessed by direct error counting (∼200 000 samples, convertedinto Q-factor).

IV. NUMERICAL RESULTS AND DISCUSSIONS

A. Back-to-Back Transmission

In this section, we aim to establish maximum transmittablebaud rates for various modulation formats, for a given absolute

Fig. 3. Q-factor as a function of baud rate for PM-mQAM signals, both with-out (red circles) and with (blue squares) digital pre-emphasis. (a) PM-4QAM,(b) PM-8QAM, (c) PM-16QAM. Inset shows the normalized Q-penalties.

performance penalty. Fig. 3 shows the Q-factor as a function ofincreasing baud rate, both without and with digital pre-emphasis,assuming a typical DAC specification (16 GHz, 5.5 bits). Alsoshown in the figures is the theoretical performance degradation,assuming no filtering and quantization effects.

RAFIQUE et al.: DIGITAL PRE-EMPHASIS IN OPTICAL COMMUNICATION SYSTEMS: ON THE NONLINEAR PERFORMANCE 143

TABLE ISUMMARY OF BAUD RATES FROM FIG. 3

Modulation ∗9 dB Q-factor ∗∗8 dB Q-factor

(5.5 bits, 16 GHz) (5.5 bits, 16 GHz) (5.5 bits, 16 GHz) (5.5 bits, 16 GHz)DPE OFF DPE ON DPE OFF DPE ON

PM-QPSK 40 44 44 50PM-8QAM 37 40 42 46PM-16QAM 34 38 40 44

∗Maximum ∼0.5 dB bound on device induced penalties, normalized to theory and B2Bimplementation. (0.6 dB, 0.05 dB), (0.52 dB, 0.14 dB) and (0.26 dB, 0.06 dB) for PM-4QAM, PM-8QAM, and PM-16QAM, respectively—(DPE OFF, DPE ON).∗∗ Maximum ∼1 dB bound on device induced penalties, normalized to theory and B2Bimplementation. (1.15 dB, 0.44), (1.1 dB, 0.32 dB) and (0.71 dB, 0.2 dB) for PM-4QAM,PM-8QAM, and PM-16QAM, respectively—(DPE OFF, DPE ON).

It can be seen that for any given modulation order, doublingthe baud rate, in theoretical sense, degrades the performance by∼3 dB (30 to 62 Gbaud), due to higher optical signal-to-noiseratio (OSNR) requirements with increasing baud rates. Overall,numerous observations can be ascertained from Fig. 3. (1) Theabsolute performance, with or without DPE, is worst for higherorder formats. For instance, for the lowest considered baud rate(30 Gbaud), system performance degrades, by ∼2 dB withoutDPE and ∼1 dB with DPE, compared to ideal channel response.This is mainly attributed to increasing hardware implementationpenalties, with respect to theory, with increasing modulationorder. (2) Increasing baud rate degrades performance due tolimited DAC BW and ENoB. This degradation is slightly morepronounced for higher order modulation formats, compared tolower order formats, due to multi-level drive signals—see in-sets. (3) DPE partially addresses the DAC limitations, and that,at a given absolute performance penalty, PM-4QAM achievesthe maximum gains from DPE, followed by PM-8QAM andPM-16QAM, respectively. Note that as the baud rate increases,the net performance without DPE significantly diverges fromthe theoretical curve. Indeed this performance deviation is re-duced by the DPE filter, however the degradation remains dueto substantial information loss at higher baud rates, owing tolimited DAC bandwidth of 16 GHz.

In Table I, we summarize the maximum transmittable baudrates at absolute Q-factor values of 9 and 8 dB, correspondingto �0.5 dB and �1 dB normalized Q-penalties, respectively,due to bandwidth and quantization limitation itself. As men-tioned earlier, for lower-order formats, the lower implementa-tion penalty combined with low-level drive, allows to tolerategreater signal distortion, enabling higher baud rate transmission.In particular, at 9 dB (8 dB) Q-factor, maximum baud rates em-ploying DPE are found to be 44 Gbaud (50 Gbaud), 40 Gbaud(46 Gbaud) and 38 Gbaud (44 Gbaud) for PM-4QAM, PM-8QAM and PM-16QAM, respectively. For the rest of the paper,we employ the baud rates established in this section, allowingup to ∼0.5 and ∼1 dB Q-penalties.

B. Link Transmission

Having established the maximum potential transmission ratesenabled by DPE, hereafter we investigate the impact of DPE onnonlinear transmission performance. Fig. 4 shows the Q-factor

Fig. 4. Q-factor as a function of launch power per channel for PM-16QAMWDM transmission after 400 km. (a) 38 Gbaud, (b) 44 Gbaud. Circles: DPEOFF, Squares: DPE ON.

after 400 km WDM transmission of PM-16QAM signals, em-ploying 38 Gbaud (Fig. 4(a)) and 44 Gbaud (Fig. 4(b)) config-urations. Note that results are presented both with and withoutDPE. The general trend is consistent with well-known bell-curvebehavior, where the performance is initially limited by OSNR,up to the nonlinear threshold (NLT)—at optimum launch power,beyond which channel nonlinearities limit the transmission per-formance. As expected, performance without DPE is worse,compared to the case with DPE. However, quite interestingly,it can also be observed that, compared to the linear transmis-sion (or back-to-back configuration), DPE enabled performanceimprovements are greater at the optimum launch power. In par-ticular, one may observe ∼1.3 and ∼2 dB improvements, for38 and 44 Gbaud systems, after nonlinear transmission, com-pared to ∼0.6 and ∼1 dB improvement in back-to-back trans-mission, respectively, as shown in Fig. 3(c). Consequently, the

144 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 33, NO. 1, JANUARY 1, 2015

Fig. 5. Q-factor as a function of transmission reach. (a) Diamond: 44 Gbaud PM4QAM, Triangle: 40 Gbaud PM8QAM, Circle: 38 Gbaud PM16QAM.(b) Diamond: 50 Gbaud PM4QAM, Triangle: 46 Gbaud PM8QAM, Circle: 44 Gbaud PM16QAM. Open: without DPE, Solid: with DPE. Horizontal linesrepresent pre-FEC Q-threshold (see text). Each data point at optimum launch power.

additional DPE improvements at optimum launch power inFig. 4 are attributed to mitigation of channel nonlinearities. Notethat performance similar to back-to-back configuration may alsobe observed in the optical signal to noise ratio limited lowerlaunch power region (e.g., −8 dBm) in Fig. 4. As discussed,this trend may be attributed to channel nonlinearity mitigationby the DPE filter, owing to enhanced higher frequency compo-nents by the DPE filter. This effect will be further discussed inupcoming paragraphs.

In order to confirm the linear and nonlinear performanceimprovements available from DPE across transmission reach,next we establish the maximum attainable distance by differ-ent modulation formats at baud rates established in Section IVA. Note that for various transmittable baud rates, we assumed28 Gbaud of data, including overhead, and allocated the addi-tional baud rate to increase the FEC overhead (FEC-OH). FEC-OH was calculated assuming soft-decision codes, allowing fornet coding gain equivalent to hard-decision Shannon limit [46].In particular, the FEC-OH corresponding to the used baud ratesare: Fig. 5(a): 44 Gbaud (64%), 40 Gbaud (48%), and 38 Gbaud(40%) for PM-4QAM→PM-16QAM, respectively. Fig. 5(b):50 Gbaud (88%), 46 Gbaud (72%), and 44 Gbaud (64%) for PM-4QAM→PM-16QAM, respectively. The pre-FEC Q-thresholds

are highlighted in the figure with solid horizontal lines. It isworth mentioning that FEC-OH beyond 50–60% may not bepractical today due to hardware limitations, however we presentup to 88% of OH to show the potential of our algorithm.

Fig. 5(a) shows the Q-factor as a function of transmissionreach, where each data-point is taken at NLT, from plots similarto Fig. 4. It can be seen that the maximum distance, without DPE,that could be traversed by PM-4QAM, PM-8QAM and PM-16QAM channels, at pre-FEC threshold, is ∼8600, ∼4100 and∼2400 km, respectively. Furthermore, when DPE is employed,the reported distance may be extended to ∼10000, ∼5000 and∼2900 km, respectively. These results confirm the twofold ad-vantage of our simple DPE approach. (1) The increased baudrate allows for higher FEC-OH, and lower pre-FEC threshold.(2) DPE partially mitigates channel nonlinearities across anygiven distance.

The presented results are of utmost importance and confirmthat given up to ∼0.5 dB DAC induced penalties are accepted(see Table I), the FEC-OH may be adequately enhanced, al-lowing to bridge long-haul and ultra-long-haul distances foradvanced modulation formats.

Fig. 5(b) shows results similar to Fig. 5(a), however for higherbaud rates, allowing up to 1 dB of back-to-back Q-penalty due to

RAFIQUE et al.: DIGITAL PRE-EMPHASIS IN OPTICAL COMMUNICATION SYSTEMS: ON THE NONLINEAR PERFORMANCE 145

TABLE IIABSOLUTE REACH [KM]

Modulation Fig. 5(a) Fig. 5(b)

Reach @ Pre-FEC Reach @ Pre-FEC Reach @ Pre-FEC Reach @ Pre-FECBER (DPE,OFF) BER (DPE,ON) BER (DPE,OFF) BER (DPE,ON)

PM-QPSK 8608 9976 8080 9720PM-8QAM 4160 5008 4752 5920PM-16QAM 2424 2928 2896 3680

TABLE IIIREACH IMPROVEMENT [%]

Modulation Fig. 5(a) Fig. 5(b)

Reach Impr. @ Pre-FEC BER REACH Impr. @ Pre-FEC BER

PM-QPSK 15% 21%PM-8QAM 19% 25%PM-16QAM 21% 27%

DAC limitations (see Table I). Largely, similar conclusions canbe ascertained from this figure, except that higher FEC-OHsare possible due to higher baud rates, lowering the pre-FECBER. However, one of the interesting observations is that whilePM-16QAM and PM-8QAM allow for longer reach comparedto Fig. 5(a), PM-4QAM shows up to 500 km reduced reach,compared to Fig. 5(a). This trend can be explained consider-ing that, at 50 Gbaud transmission, even though the relativeQ-penalty from DAC is limited to 1 dB at most, the net perfor-mance is degraded due to theoretical penalties associated withhigh baud rates. This phenomenon was recently studied in [47]–[48], where it was shown that the optimum baud rates may beidentified based on the tradeoff between net performance reduc-tion due to increasing overhead (or baud rate) and performancegain due to corresponding FEC-OH.

Table II summarizes the maximum transmittable reach fordifferent configurations of Fig. 5. It can be seen that the max-imum transportable distances for PM-4QAM, PM-8QAM, andPM-16QAM are ∼10000, 5900, and ∼3700 km, respectively.This result shows the potential of using currently commercialcomponents for long haul and ultra long haul transmission ofadvanced higher level modulation formats. Here, it is worthmentioning that long haul transmission for higher order formatshave been previously reported, however with the aid of advancedtransmission technologies, e.g., including Raman amplification[49], [50], digital nonlinear mitigation [40], advanced DSP [51],high effective area dispersion fibers [52], [53], etc.

Table III highlights the reach improvements at pre-FECthreshold, relative to the configuration without DPE. As dis-cussed earlier, greater improvements are available for higherbaud rate systems (from Fig. 5(b)). Moreover, the reach im-provement increase with increasing modulation order. Note that,consistent with Fig. 3, the net improvement in reach is stillgreater for lower order formats, however the relative percentageincrease is greater for higher order formats due to relativelyshorter traversed distance.

Fig. 6. Q-improvement as a function of transmission reach. (a) Di-amond: 44 Gbaud PM4QAM, Triangle: 40 Gbaud PM8QAM, Circle:38 Gbaud PM16QAM. (b) Diamond: 50 Gbaud PM4QAM, Triangle: 46 GbaudPM8QAM, Circle: 44 Gbaud PM16QAM. Inset: Back-to-Back performanceimprovements. Each data point at optimum power.

Having established maximum transmission reach at pre-FECthreshold for various configurations, in Fig. 6 we explore theDPE dependency on distance, and plot the Q-improvements,enabled by DPE, as a function of transmission reach. It can beclearly seen that DPE enables greater improvements at shorterreach, and the performance benefits saturate as longer linksare traversed. In particular, in Fig. 6(a), PM-16QAM enablesmaximum improvement of up to ∼1.6 dB, which saturates to∼0.8 dB after ∼11 spans. On the other hand, PM-8QAM andPM-4QAM almost show no dependency on reach due to highertraversed reach, essentially higher accumulated dispersion dom-inating the nonlinearity averaging effects due to rapidly broad-ening pulses [54], [55]. However, in Fig. 6(b), this trend is morepronounced, and clear improvements can be seen for trans-mission below 11 spans, irrespective of the modulation order.Also note that, similar to Fig. 4, even after DPE gain satura-tion, the Q-improvement after nonlinear transmission is greaterthan in back-to-back transmission (see insets). These resultsmay be attributed to channel nonlinearity compensation by theDPE filter, where higher frequencies are enhanced to mitigatelow-pass DAC impact. In time-domain this translates to rapidlydispersing pulses, allowing for better averaged intra-channelnonlinear effects and larger walk off for inter-channel nonlinear-ities. Here it is worth mentioning that, for shorter links without

146 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 33, NO. 1, JANUARY 1, 2015

DPE, intra-channel nonlinearities are only partially mitigateddue to limited accumulated dispersion, allowing for larger netQ-improvements from DPE. On the other hand, for longer links,even though DPE enables performance improvements in initialspans, similar to shorter links, the higher accumulated disper-sion leads to substantial intra-channel nonlinear mitigation evenwithout DPE, leading to reduced net Q-improvement in thiscase. In other words, in case of dispersion unmanaged long haultransmission, nonlinear averaging may be intrinsically achievedwithout DPE, however for shorter distances (<∼1000 km) suchaveraging is largely achieved through higher frequency enhance-ment by the DPE filter. Nonetheless, the overall performance isworst for longer links due to reduced OSNR and additional perspan inter-channel nonlinearity contributions.

These results provide an interesting and practical outcome ofDPE, confirming that transmitter based DPE not only mitigatesDAC induced signal distortion, but also enables better nonlinearchannel performance. Using Figs. 5 and 6, performance regimesfor different modulation and baud rates may be established,where gain of DPE in terms of higher FEC-OH and nonlinearitycompensation may be recognized.

Next, we investigate the impact of channel spacing on the per-formance improvements enabled by DPE algorithm, for all themodulation formats and baud rates, allowing up to 0.5 and 1 dBpenalties, as established in Table I. Fig. 7(a) shows Q-factor as afunction of channel spacing, targeting reach at BER of 10−3 (Q-factor of ∼9.8 dB) for PM-4QAM (42spans). It can be clearlyseen that the minimum required channel spacing is ∼1.1 ×BaudRate, and that beyond this saturation point, transmissionperformance, both with and without DPE, is independent ofchannel spacing. However, when channel spacing is below 1.1× Baud rate, the performance is severely limited by linear andnonlinear channel crosstalk. In this case, the performance im-provements from DPE are diminished, confirming that DPE isnot very effective to compensate inter-channel nonlinearities,and that our algorithm does not work well in case of high linearinter-carrier crosstalk. Furthermore, similar trends are seen forboth high and low baud rate transmission regimes.

Fig. 7(b) and (c) show results similar to Fig. 7(a), for PM-8QAM and PM-16QAM, respectively. It is clear that the mini-mum channel spacing requirement is increased to ∼1.15 × BRand ∼1.2 × BR, however the DPE performance trend is simi-lar to Fig. 7(a), where performance improvement is decreasedbelow the minimum saturation point.

The above analysis confirms that as far as channel nonlin-earity mitigation is concerned, simple DPE filter is unable tomitigate inter-channel fiber nonlinearities, however it workswell for intra-channel nonlinear mitigation. This is consistentwith our earlier hypothesis that higher frequency enhancementleads to increased pulse spread, allowing for better intra-channelnonlinear averaging. Recently, this phenomenon was also ob-served in context of special pulse-design, termed as M-shapedpulses [56], which further verifies our findings. Nonetheless,our approach of DPE enables the nonlinear suppression at noadditional hardware cost.

Fig. 7. Q-factor as a function of channel spacing. (a) PM-4QAM after3360 km, Circle: 44Gbaud, Square: 50 Gbaud, (b) PM-8QAM after 1360 km,Circle: 40 Gbaud, Square: 46 Gbaud, (c) PM-16QAM after 560 km, Circle:38 Gbaud, Square: 44 Gbaud. Open: without DPE, Solid: with DPE. Each datapoint at optimum power.

RAFIQUE et al.: DIGITAL PRE-EMPHASIS IN OPTICAL COMMUNICATION SYSTEMS: ON THE NONLINEAR PERFORMANCE 147

Fig. 8. Experimental transmission setup. (a) Terabit super-channel transmitter,(b) Optical spectrum of 36 GBd × 4-subcarriers PM-16QAM super-channel,(c) Re-circulating optical loop and receiver. WSS: Wavelength selective switch,PS: Polarization scrambler, BPF: Band-pass filter, SSMF: Standard Single ModeFiber.

Fig. 9. Experimentally measured frequency response of a commercial digital-to-analog converter and driver for four different ports.

V. EXPERIMENTAL TRANSMISSION CONFIGURATION

In order to confirm the nonlinear benefit of our DPE algo-rithm, in this section we experimentally verify its performance.The transmission experiment employed quad-carrier 1.15 Tb/sPM-16QAM WDM transmission (36 Gbaud per carrier) [30].

As shown in Fig. 8, the transmit data was first mapped to16QAM symbols followed by up-sampling and RRC filter-ing with a roll-off factor of 0.2 to limit bandwidth and mini-mize inter-carrier crosstalk. Using the experimentally measuredjoint transfer function of the DAC and driver (see Fig. 9), i.e.,DDAC,REAL (f), we applied pre-distortion of low pass filteringeffects using algorithm defined in Section II. Note that we scaled

the calculated P (f) with an additional ramp to compensate forIQ-modulator effects. The slope of the ramp was optimized bymonitoring the BER at system startup.

A pair of IQ-modulators were used to modulate even andodd carriers of four-subcarrier terabit super-channel. Each IQ-modulator was fed by two integrated tunable laser assemblies,coupled together using a polarization maintaining coupler toretain the polarization properties. After modulation, both pairsof subcarriers went through polarization multiplex emulation(PME) stage where a single polarization signal was split intwo orthogonal polarizations by a polarization beam splitterand then the two orthogonal polarizations were combined by apolarization beam combiner after one of them had been delayed.After PME, the even and odd subcarriers were combined by a3 dB coupler to generate a four subcarrier super-channel, asshown in Fig. 8(c).

The subcarrier spacing was set to 1.2 × baud rate, and thequad-carrier terabit super-channel was placed in 200 GHz grid.Note that we placed the Terabit super-channel in a 200 GHzgrid, in order to enable practical flex-grid scenario assumingspectral buffer to accommodate filtering penalties.

The transmission setup consisted of 8 × 100 G PM-4QAMsignals evenly distributed to the left and right of terabit super-channel in a 50 GHz ITU grid. The WDM signal was then am-plified by an EDFA before going through a fast optical switchbased on variable optical attenuators. After passing through theswitch, the optical signal was passed through a re-circulatingoptical loop which consisted of four spans of 95 km SSMFfollowed by EDFA. In addition to fiber and EDFAs, a loop syn-chronous polarization scrambler was placed at the beginningof the loop to eliminate periodic signal conditioning by loopcomponents. Also a wavelength selective switch was employedto minimize the out of band noise. After re-circulating in theloop N times, the optical signal was filtered by a 0.6 nm tunableband-pass filter. The optical signal was then mixed with the localoscillator in the optical front-end, sampled by a 50 GSa/s oscillo-scope, and finally saved for offline processing. The results wereevaluated using second subcarrier of the terabit super-channel(centered at 1550.15 nm), as it showed the worst performance(linear plus nonlinear crosstalk).

VI. EXPERIMENTAL RESULTS AND DISCUSSIONS

Fig. 10 shows the measured Q-factor (converted from BER)as a function of transmission reach for 36 Gbaud quad-carrier1.15 Tb/s PM-16QAM super-channel, co-propagating with100 Gb/s PM-QPSK neighbors. The optimum launch powerwas determined by a single span transmission of super-channelincluding the WDM neighbors, found to be ∼1.5 dBm per chan-nel. Moreover, the optimum launch power remained the samewith or without DPE.

As expected, DPE enables longer transmission reach, com-pared to the case without DPE. Moreover, consistent withnumerical results, greater improvements are available afternonlinear transmission, compared to back-to-back transmis-sion, and that DPE gain increases with decreasing span count.In particular, it can be seen that∼1.1 dB Q-improvement may beattained after ∼380 km, and that DPE gain reduces to ∼0.9 dB

148 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 33, NO. 1, JANUARY 1, 2015

Fig. 10. Q-factor as a function of transmission reach for 36 Gbaud quad-carrier 1.15 Tb/s PM-16QAM WDM transmission over SSMF. Open circle:without DPE, solid circle: with DPE. Each data point at optimum launch power.

at 740 km. These experimental results confirm the conclusionsdrawn from our exhaustive simulation results that DPE filter en-ables channel nonlinearity mitigation, in addition to DAC lowpass response cancellation. Note that, compared to the earlierpresented simulative analysis, DPE induced nonlinear perfor-mance improvements are much lower in experimental settings.This can be explained considering that DPE filter tends to mit-igate intra-channel nonlinearities, with limited improvement incase of higher inter-channel crosstalk, as shown in Fig. 7. In thiscontext, even though terabit super-channel sub-carriers werespaced at ∼1.2 × baud rate, the LASERs were not frequencylocked, leading to inter-channel crosstalk, essentially limitingDPE gains. Nonetheless, this effect can be overcome eitherby increasing the subcarrier spacing or using frequency lockedLASERs.

VII. CONCLUSION

We conducted a comprehensive investigation on nonlin-ear performance improvements enabled by a simple digitalpre-emphasis algorithm, selectively enhancing higher signalfrequencies. DPE is shown to enable two-fold advantages, (1) In-creasing maximum potential transmission rates using currentlycommercial components, (2) Improving nonlinear channel toler-ance. In particular, we showed maximum baud rates of up to 50,46, and 44 Gbaud for PM-4QAM→PM-16QAM, respectively.Furthermore, we showed, for the first time to our knowledge,that DPE mitigates intra-channel nonlinearities, with greater im-provements for higher order formats and lower span count. Weverified our results in both numerical and experimental configu-rations, and confirmed that DPE induced nonlinear performanceimprovement may be attained without using any additionalhardware.

Our results suggest that next-generation transponders em-ploying DPE will benefit from nonlinear performance gains,which should be taken into account for system margin alloca-tions, in addition to more conventional improved performancedue to DAC pre-emphasis.

REFERENCES

[1] C. R. S. Fludger, T. Duthel, D. Borne, C. Schulien, E. Schmidt, T. Wuth,J. Geyer, E. Man, G. Khoe, and H. Waardt, “Coherent equalization and

POLMUX-RZ-DQPSK for robust 100-GE transmission,” J. Lightw. Tech-nol., vol. 26, no. 1, pp. 64–72, Jan. 1, 2008.

[2] (2014, May). Coriant 100G Product Datasheet. [Online]. Available: http://coriant.com/products/hit7300.asp.

[3] Z. Dong, X. Li, J. Yu, and N. Chi, “6× 128-Gb/s Nyquist-WDM PDM-16QAM generation and transmission over 1200-km SMF-28 with a SEof 7.47 b/s/Hz,” J. Lightw. Technol., vol. 30, no. 24, pp. 4000–4005,Dec. 2012.

[4] M. Huang, A. Tanaka, E. Ip, Y. Huang, D. Qian, Y. Zhang, S. Zhang,P. N. Ji, I. B. Djordjevic, T. Wang, Y. Aono, S. Murakami, T. Tajima,T. J. Xia, and G. A. Wellbrock, “Terabit/s Nyquist superchannels in highcapacity fiber field trials using DP-16QAM and DP-8QAM modulationfor regional and long-haul applications,” J. Lightw. Technol., vol. 32,no. 4, pp. 776–782, Feb. 2014.

[5] D. Rafique, T. Rahman, A. Napoli, S. Calabro, and B. Spinnler, “Technol-ogy options for 400Gb/s PM-16QAM flex-grid network upgrades,” IEEEPhoton. Technol. Lett., vol. 26, no. 8, pp. 773–776, Apr. 2014.

[6] I. Stiakogiannakis, E. Palkopoulou, D. Klonidis, O. Gerstel, and I. Tomkos,“Dynamic cooperative spectrum sharing and defragmentation for elasticoptical networks,” J. Opt. Commun. Netw. vol. 6, no. 3, pp. 259–269,Feb. 2014.

[7] G. Bosco, V. Curri, A. Carena, P. Piggiolini, and F. Forghieri, “On theperformance of Nyquist-WDM terabit superchannels based on PM-BPSK,PM-QPSK, PM-8QAM, or PM-16QAM subcarriers,” J. Lightw. Technol.,vol. 29, no. 1, pp. 53–61, Jan. 2011.

[8] M. Ruiz, L. Velasco, A. Lord, D. Fonseca, M. Pioro, R. Wessaly, andJ. P. Fernandez-Palacios, “Planning fixed to flexgrid gradual migration:drivers and open issues,” IEEE Commun. Mag., vol. 52, no. 1, pp. 70–76,Jan. 2014.

[9] D. Rafique, T. Rahman, A. Napoli, M. Kuschnerov, and B. Spinnler,“Flex-grid optical networks: Spectrum allocation and nonlinear dynam-ics of super-channels,” Opt. Exp., vol. 21, no. 26, pp. 32184–32191,Dec. 2013.

[10] L. Zong, G. N. Liu, A. Lord, Y. R. Zhou, and T. Ma, “40/100/400 Gb/smixed line rate transmission performance in flexgrid optical networks,”presented at the Opt. Fiber Commun. Conf., Anaheim, CA, USA, 2013,Paper OTu2A.2.

[11] C. Laperle, M. O. Sullivan, and K. Roberts, “Advances in high-speedADCs, DACs, and DSP for optical transceivers acknowledgement,” J.Lightw. Technol., vol. 32, no. 4, pp. 629–643, Feb. 2014.

[12] M. Nolle, F. Frey, R. Elschner, C. Schmidt-Langhorst, A. Napoli, andC. Schubert, “Performance comparison of different 8QAM constellationsfor the use in flexible optical networks,” presented at the Opt. Fiber Com-mun. Conf., 2014, San Francisco, CA, USA, Paper W3B.2.

[13] R. I. Killey, P. M. Watts, M. Glick, and P. Bayvel, “Electronic disper-sion compensation by signal predistortion,” presented at the Opt. FiberCommun. Conf., Anaheim, CA, USA, 2006.

[14] T. Oyama, H. Nakashima, T. Hoshida, Z. Tao, C. Ohshima, andJ. C. Rasmussen, “Efficient transmitter-side nonlinear equalizer for 16QAM,” presented at the Eur. Conf. Opt. Commun., London, U.K., 2013,Paper We.3.C.1.

[15] Tektronix Digital Oscilloscope Datasheets. (2014, May). [Online]. Avail-able: http://www.tek.com/oscilloscope#all.

[16] Lecroy digital scope datasheets. (2014, May). [Online]. Available: http://teledynelecroy.com/oscilloscope/.

[17] Fujitsu DAC datasheets. (2014, May). [Online]. Available: http://www.fujitsu.com/emea/services/microelectronics/dataconverters/chais/.

[18] J. Wang, C. Xie, S. Member, and Z. Pan, “Optimization of DSP to gen-erate spectrally efficient 16 QAM nyquist-WDM signals,” IEEE Photon.Technol. Lett., vol. 25, no. 8, pp. 772–775, Apr. 2013.

[19] Z. Jia, H.-C. Chien, J. Zhang, Z. Dong, and Y. Cai, “Performance analysisof pre- and post-compensation for bandwidth-constrained signal in high-spectral-efficiency optical coherent systems,” presented at the Opt. FiberCommun. Conf., San Francisco, CA, USA, 2014, Paper W3K.3.

[20] J. Zhang, H. Chien, Y. Xia, Y. Chen, and J. Xiao, “A novel adaptive digitalpre-equalization scheme for bandwidth limited optical coherent systemwith DAC for signal generation,” presented at the Opt. Fiber Commun.Conf., San Francisco, CA, USA, 2014, Paper W3K.4.

[21] C. Rasmussen, Y. Pan, M. Aydinlik, M. Crowley, J. Geyer, P. Humblet,F. Liu, B. Mikkelsen, P. Monsen, N. Nadarajah, G. Pendock, and B. Shah,“Real-time DSP for 100+Gb/s,” presented at the Opt. Fiber Commun.Conf., Anaheim, CA, USA, 2013, Paper OW1E.1.

[22] A. Ghazisaeidi, P. Tran, P. Brindel, O. Bertran-pardo, J. Renaudier,G. Charlet, and S. Bigo, “Impact of tight optical filtering onthe performance of 28 Gbaud Nyquist-WDM PDM-8QAM over37. 5 GHz grid,” presented at the Opt. Fiber Commun. Conf., 2013, PaperOTu3B.6.

RAFIQUE et al.: DIGITAL PRE-EMPHASIS IN OPTICAL COMMUNICATION SYSTEMS: ON THE NONLINEAR PERFORMANCE 149

[23] C. E. Tibbs, “A review of wide band frequency modulation technique,”J. Brit. Inst. Radio Eng. , vol. 4, no. 2, p. 85, Jun.–Sep. 1944.

[24] R. Farjad-rad, C.-K. K. Yang, M. A. Horowitz, and T. H. Lee, “A 0.4-μmCMOS 10-Gb/s 4-PAM pre-emphasis serial link transmitter,” IEEEJ. Solid-State Circuits, vol. 34, no. 5, pp. 580–585, May 1999.

[25] J. Renaudier, O. Bertran-Pardo, H. Mardoyan, P. Tran, G. Charlet, S. Bigo,A. Konczykowska, J.-I. Dupuy, F. Jorge, M. Riet, and J. Godin, “Spectrallyefficient long-haul transmission of 22-Tb/s using 40-Gbaud PDM-16QAMwith coherent detection,” presented at the Opt. Fiber Commun. Conf.,2012, Paper OW4C.2.

[26] T. Sugihara, T. Kobayashi, T. Fujimori, and T. Mizuochi, “Electronic pre-equalization technologies using high-speed DAC,” presented at the Eur.Conf. Opt. Commun., Geneva, Switzerland, 2011, Paper Tu.6.B.2.

[27] W. Idler, F. Buchali, K. Schuh, D. Roesener, E. Lach, and A. Leven,“Spectral pre-distortion with FPGA and DAC at 448-Gb/s DP-16QAMimproving nonlinear threshold power,” presented at the Opt. Fiber Com-mun. Conf., Anaheim, CA, USA, 2013, Paper OTh3C.1.

[28] J. Qi, B. Mao, N. Gonzalez, L. N. Binh, and N. Stojanovic, “Generation of28GBaud and 32GBaud PDM-Nyquist-QPSK by a DAC with 11.3 GHzanalog bandwidth,” presented at the Opt. Fiber Commun. Conf., Anaheim,CA, USA, 2013, Paper OTh1F.1.

[29] Z. Dong, H. C. Chien, Z. Jia, and X. Li, “Joint digital preequalizationfor spectrally efficient super Nyquist-WDM signal,” J. Lightw. Technol.,vol. 31, no. 20, pp. 3237–3242, 2013.

[30] T. Rahman, D. Rafique, A. Napoli, E. de Man, M. Kuschnerov, B. Spinnler,M. Bohn, C. Okonkwo, and H. de Waardt, “Long-haul terabit transmission(2272 km) employing digitally pre-distorted quad-carrier PM-16QAMsuper-channel,” presented at the Eur. Conf. Opt. Commun., 2014, PaperWe.1.3.5.

[31] C. Behrens, S. Member, R. I. Killey, S. J. Savory, M. Chen, and P. Bayvel,“Nonlinear transmission performance of higher-order modulation for-mats,” IEEE Photon. Technol. Lett., vol. 23, no. 6, pp. 377–379,Apr. 2011.

[32] A. D. Ellis, D. Rafique, and S. Sygletos, “Long-haul terabit trans-mission (2272 km) employing digitally pre-distorted quad-carrier PM-16QAM super-channel” in Proc. Eur. Conf. Opt. Commun., 2011,Paper TuN1.

[33] E. Ip, “Nonlinear compensation using backpropagation for polarization-multiplexed transmission,” J. Lightw. Technol., vol. 28, no. 6, pp. 939–951,2010.

[34] D. Rafique and A. D. Ellis, “Scaling the benefits of digital nonlinear com-pensation in high bit-rate optical meshed networks,” in Current Develop.Opt. Fiber Technol., Rijeka, Croatia: InTech, 2013, pp. 21–42.

[35] M. Morshed, L. B. Du, and A. J. Lowery, “Mid-span spectral inversionfor coherent optical OFDM systems: fundamental limits to performance,”J. Lightw. Technol., vol. 31, no. 1, pp. 58–66, Jan. 2013.

[36] D. Rafique, T. Rahman, A. Napoli, and B. Spinnler, “Intra super-channelfiber nonlinearity compensation in flex-grid optical networks,” Opt. Exp.,vol. 21, no. 26, pp. 32063–32070, Dec. 2013.

[37] S. Oda, T. Hoshida, H. Nakashima, Y. Aoki, Z. Tao, and J. C. Rasmussen,“Impact of perturbation back-propagation on carrier phase recovery in224 Gb/s DP-16QAM transmission,” presented at the Optoelectron. Com-mun. Conf., Kyoto, Japan, 2013, Paper WR4–1.

[38] Y. Gao, J. C. Cartledge, A. S. Karar, S. S. Yam, M. O. Sullivan, C. Laperle,A. Borowiec, and K. Roberts, “Reducing the complexity of perturbationbased nonlinearity pre-compensation using symmetric EDC and pulseshaping,” Opt. Exp., vol. 22, no. 2, pp. 1209–1219, Jan. 2014.

[39] E. Yamazaki, M. Tomizawa, and Y. Miyamoto, “100-Gb/s optical transportnetwork and beyond employing digital signal processing,” IEEE Commun.Mag., vol. 50, no. 2, pp. 43–49, Feb. 2012.

[40] L. B. Du, D. Rafique, A. Napoli, B. Spinnler, A. D. Ellis, M. Kuschnerov,and A. J. Lowery, “Digital fiber nonlinearity compensation: Towards1-Tb/s transport,” IEEE Signal Process. Mag., vol. 31, no. 2, pp. 46–56,Mar. 2014.

[41] A. Napoli, Z. Maalej, V. A. J. M. Sleiffer, M. Kuschnerov, D. Rafique,E. Timmers, B. Spinnler, T. Rahman, L. D. Coelho, and N. Hanik, “Re-duced complexity digital back-propagation methods for optical commu-nication systems,” J. Lightw. Technol., vol. 32, no. 7, pp. 1351–1362,Apr. 2014.

[42] D. Rafique, A. Napoli, S. Calabro, and B. Spinnler, “Digital pre-emphasison optical communication systems: on the DAC requirements for ter-abit transmission applications,” J. Lightw. Technol., vol. 32, no. 19,pp. 3247–3256, Jul. 2014.

[43] T. Oyama, T. Hoshida, H. Nakashima, C. Ohshima, Z. Tao, andJ. C. Rasmussen, “Impact of pulse shaping and transceiver electrical band-

widths on nonlinear compensated transmission,” presented at the Opt.Fiber Commun. Conf., Anaheim, CA, USA, 2013, Paper OTh3C.2.

[44] J.-X. Cai, Y. Cai, Y. Sun, C. R. Davidson, D. G. Foursa, A. Lucero,O. Sinkin, W. Patterson, A. Pilipetskii, G. Mohs, and N. Bergano, “112 ×112 Gb/s transmission over 9,360 km with channel spacing set to thebaud rate (360% spectral efficiency),” presented at the Eur. Conf. Opt.Commun., Torino, Italy, 2010, Paper PD2.1.

[45] M. Kuschnerov, F. N. Hauske, K. Piyawanno, B. Spinnler, M. S. Alfiad,A. Napoli, and B. Lankl, “DSP for coherent single-carrier receivers,”J. Lightw. Technol., vol. 27, no. 16, pp. 3614–3622, Feb. 2009.

[46] L. Schmalen and A. J. D. L. Van Wijngaarden, “Forward error correctionin optical core and optical access networks,” Bell Labs Tech. J., vol. 18,no. 3, pp. 39–66, Nov. 2013.

[47] T. Rahman, D. Rafique, A. Napoli, E. de Man, M. Kuschnerov, B. Spinnler,M. Bohn, C. Okonkwo, and H. de Waardt, “FEC overhead optimizationfor long-haul transmission of PM-16QAM based 400Gb/s super-channel,”presented at the Eur. Conf. Opt. Commun., Cannes, France, 2014,Paper P.5.17.

[48] R. Rios-Muller, J. Renaudier, P. Brindel, C. Simonneau, P. Tran,A. Ghazisaeidi, I. Fernandez, L. Schmalen, and G. Charlet, “Opti-mized spectrally efficient transceiver for 400-Gb/s sinlge carrier trans-port,” presented at the Eur. Conf. Opt. Commun., Cannes, France, 2014,Paper PD.4.2.

[49] L. E. Nelson, S. Member, X. Zhou, B. Zhu, M. F. Yan, P. W. Wisk, andP. D. Magill, “All-Raman-amplified, 73 nm seamless band transmissionof 9 Tb/s over 6000 km of fiber,” IEEE Photon. Technol. Lett., vol. 26,no. 3, pp. 242–245, Feb. 2014.

[50] T. J. Xia, G. A. Wellbrock, D. Chang, S. Burtsev, W. Pelouch, E. Zak,H. De Pedro, W. Szeto, and H. Fevrier, “Transmission of 400G PM-16QAM channels over long-haul distance with commercial all-distributedRaman amplification system and aged standard SMF in field,” presentedat the Opt. Fiber Commun. Conf. Exhib., San Francisco, CA, USA, 2014,Paper Tu2B.1.

[51] X. Liu, S. Chandrasekhar, and P. J. Winzer, “Fiber-nonlinearity-tolerantsuperchannel transmission via nonlinear noise squeezing and general-ized phase-conjugated twin waves,” J. Lightw. Technol., vol. 32, no. 4,pp. 766–775, Feb. 2014.

[52] X. Liu, S. Chandrasekhar, P. J. Winzer, T. Lotz, J. Carlson, J. Yang,and G. Cheren, “1.5-Tb/s guard-banded superchannel transmission over56×100 km (5600-km) ULAF using 30-Gbaud pilot-free OFDM-16QAM signals with 5.75-b/s/Hz net spectral efficiency,” presented atthe Eur. Conf. Opt. Commun., Amsterdam The Netherlands , 2012,Paper PDPTh3.C.5.

[53] V. A. J. M. Sleiffer, Z. Maalej, D. van den Borne, M. Kuschnerov,V. Veljanovski, M. Hirano, Y. Yamamoto, T. Sasaki, S. L. Jansen,A. Napoli, and H. de Waardt, “A comparison between SSMF and large-Aeff pure-silica core fiber for ultra long-haul 100G transmission,” Opt.Exp., vol. 19, no. 26, pp. B710–B715, Dec. 2011.

[54] G. P. Agrawal, Nonlinear Fiber Optics, 4th ed. New York, NY, USA:Academic, 2007.

[55] C. R. Menyuk and B. S. Marks, “Interaction of polarization mode dis-persion and nonlinearity in optical fiber transmission systems,” J. Lightw.Technol., vol. 24, no. 7, pp. 2806–2826, Jul. 2006.

[56] X. Xu, Q. Zhuge, B. Chatelain, M. Qiu, M. Chagnon, M. Morsy-Osman,W. Wang, and D. V. Plant, “Nonlinearity-tolerant frequency domain rootM-shaped pulse for spectrally efficient coherent transmissions,” presentedat the Opt. Fiber Commun. Conf., San Francisco, CA, USA, 2014,Paper W1G.3.

Danish Rafique received the B.Sc. degree from FAST-NU, Pakistan, in 2007, theM.Sc. degree from, KTH, Sweden, in 2009, and the Ph.D. degree from TyndallNational Institute, University College Cork, Ireland, in 2012, all in electricalengineering. He joined Nokia Siemens Networks in 2012, and is currentlypursuing a career with R&D Labs of Coriant GmbH (former NSN), Germany,where he is working on product development and research programs on next-generation transponder solutions involving flex-grid networks. His interestsinclude nonlinear effects in fiber, optical system design tradeoffs, digital signalprocessing, and dynamic optical networks. He has published more than 60articles, including peer-reviewed journals, contributed and invited conferencepublications, and book chapters, and acts as a reviewer for IEEE PHOTONICS

TECHNOLOGY LETTERS, IEEE JOURNAL OF LIGHTWAVE TECHNOLOGY, OSAOptics Express, etc.

150 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 33, NO. 1, JANUARY 1, 2015

Talha Rahman was born in Lahore, Pakistan, in September 1985. He receivedthe B.Sc. degree in communication systems engineering from Institute of SpaceTechnology, Islamabad, Pakistan, in 2007 and the M.Sc.(Hons.) degree in ad-vanced optical technologies with specialization in optical communication sys-tems from FAU Erlangen-Nuernberg, Erlangen, Germany, in 2012. He con-ducted his internship and master thesis research at Alcatel-Lucent Bell LabsStuttgart Germany on coherent transmission systems with specialized 4-D mod-ulation formats. Since March 2013, he has been with Eindhoven University ofTechnology as a Ph.D. candidate under EU-FP7 project IDEALIST. He con-ducts his research at Coriant R&D labs in Munich, Germany. His researchinterests include flexible optical transceivers focusing on 400 Gb/s and 1 Tb/ssuperchannels.

Antonio Napoli received the Master’s and Ph.D. degrees in electronics engi-neering from Politecnico di Torino, Torino, Italy, in 2002 and 2006, respectively.He was a Visiting Student at the Technische Universitat Wien in 2001, a Visit-ing Researcher at Universitat Politecnica de Catalunya in Barcelona, in 2004,and at University College London, U.K., in 2005. In 2006, he joined SiemensCOM in Munich, where he worked on EDFA transient suppression and DSPfor long-haul optical networks. In 2007, he joined Nokia Siemens Networks,where he worked on optical transmission up to 400 Gb/s. He has been withCoriant since 2013, where is leading a work-package on data plane within theEU FP-7 IDEALIST project. He regularly supervises students, teaches at severaluniversities and acts as Reviewer for magazines such as IEEE/OSA JOURNAL

OF LIGHTWAVE TECHNOLOGY or IEEE PHOTONICS TECHNOLOGY LETTERS. Hisresearch interests include DSP and advanced modulation formats for long-haulsystems. He is a (co)author of about 80 peer-reviewed papers and conferencecontributions, and six patents.

Bernhard Spinnler received the Dipl.-Ing. degree in communications engi-neering and the Dr.-Ing. degree from the University of Erlangen-Nurnberg,Erlangen, Germany, in 1994 and 1997, respectively. Since 1997, he has beenworking on low-complexity modem design of wireless radio relay systems atSiemens AG, information and communication networks. In 2002, he joined theoptical networks group of Siemens Corporate Technology, which later becameNokia Siemens Networks. He has been with Coriant since 2013, where he isworking on robust and tolerant design of optical communications systems. Hisresearch interests include advanced modulation and equalization for single andmulticarrier systems.