high-capacity fiber field trial

546 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 4, FEBRUARY 15, 2013

High-Capacity Fiber Field Trial Using Terabit/sAll-Optical OFDM Superchannels With DP-QPSK

and DP-8QAM/DP-QPSK ModulationYue-Kai Huang, Ming-Fang Huang, Ezra Ip, Eduardo Mateo, Philip N. Ji, Dayou Qian, Akihiro Tanaka,

Yin Shao, Ting Wang, Yoshiaki Aono, Tsutomu Tajima, Tiejun J. Xia, and Glenn A. Wellbrock

AbstractWe report the results of two separate field trialsaimed to achieve high fiber capacity over long-haul distances.In the first trial, 22 all-optical orthogonal frequency-divisionmultiplexing superchannels with hybrid dual-polarization 8quadrature amplitude modulation and dual-polarization quadra-ture phase-shift-keying (DP-QPSK) modulation were generatedusing a novel flexible format superchannel transmitter design. Thesignal carrying net 21.7 Tb/s data was transmitted over 1503 kmof dispersion uncompensated field-installed standard single-modefiber with the aid of hybrid Raman and erbium-doped fiberamplifier (EDFA) amplification and digital coherent detection. Inthe second trial, we extended the transmission distance to over2531 km of field fiber using only EDFA for loss compensation.The increase in reach was achieved by reducing the net totaldata rate to 16.2 Tb/s and modulating the optical superchannelsubcarriers with DP-QPSK only. To the best of our knowledge, weachieved the highest capacity field trial record to date at 21.7 Tb/sin the first trial, while the achieved capacitydistance product of40.9 Pb/s km in the second trial is also the highest reported to date.

Index TermsCoherent communications, fiber optical commu-nications, optical signal processing, orthogonal frequency-divisionmultiplexing (OFDM).

I. INTRODUCTION

I NCREASING fiber capacity is required by network oper-ators to cope with year over year traffic growth. Current100G systems using dual-polarization quadrature phase shiftkeying (DP-QPSK) and digital coherent detection operate at achannel spacing of 50GHz, allowing network operators to attaina capacity of 9.6 Tb/s (96 100 Gb/s) per fiber using theC-bandalone [1], [2]. Upgrading channel capacity from 10 Gb/s/50GHz systems to 100 Gb/s/50 GHz systems has increased thetotal capacity per fiber by a factor of 10 [1] and by further re-ducing the spacing between optical carriers, the total capacity

Manuscript received June 01, 2012; revised August 22, 2012; acceptedSeptember 10, 2012. Date of publication October 24, 2012; date of currentversion January 07, 2013.Y.-K. Huang, M.-F. Huang, E. Ip, E. Mateo, P. N. Ji, D. Qian, A. Tanaka,

Y. Shao, and T. Wang are with NEC Laboratories America, Inc., Princeton, NJ08540 USA (e-mail: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected];[email protected]; [email protected]; [email protected]).Y. Aono and T. Tajima are with NEC Corporation, Tokyo 101-8532, Japan

(e-mail: [email protected]; [email protected]).T. J. Xia and G. A. Wellbrock are with Verizon, Richardson, TX 75082 USA

(e-mail: [email protected]; [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.2012.2226384

per fiber can be increased by another 50% [3]. By using largersized signal constellations in conjunction with more lasers overa wider transmission window, two recent experiments have re-ported capacities higher than 100 Tb/s over a single fiber core[4], [5]. However, these hero experiments require specialtyfibers which are not found in existing networks. In fact, mosthigh fiber capacity results reported so far have been obtainedwith better transmission link conditions than what is obtainablefrom a real fiber network in the field. For example, newly de-veloped transmission fiber, such as ultralarge core fiber or ul-tralow loss fiber, has played an important role in high-capacityresearch. These results have set a capacity expectation for net-work operators that could only be achieved if the new fiber wereto be deployed in the future. High fiber capacity experiments inthe field, however, are a different story. So far, reported fibercapacities in field trials are below 10 Tb/s (An example is givenin [2].). It is important to investigate how much capacity net-work operators can achieve with the deployed fiber infrastruc-ture, since costly large-scale network-wide fiber upgrades arenot expected any time soon.Superchannel design using orthogonal frequency-division

multiplexing (OFDM) is a promising transmission platformfor data rate beyond 100G to increase fiber capacity owingto the parallelized transmitter and receiver architecture andhigh spectral efficiency. Terabit-per-second transmission usingOFDM has been demonstrated [6], [7], in which an electricalOFDM signal is generated using digital-to-analog converters(DACs) and optical techniques creating identical copies fillingthe required spectrum are used. The subcarrier data rates ofsuch systems are limited by the bandwidth of available DACs,typically in the range of 10 to 15 GHz. An alternative ap-proach to creating OFDM signals at 1 Tb/s and beyond is togenerate the OFDM subcarriers optically and modulating themas single-carrier (SC) signals [3], [8][10]. The modulatedsubcarriers are combined to produce the all-optical OFDM(AO-OFDM) signal. To detect the AO-OFDM signal, wecan demultiplex the subcarriers using an optical fast Fouriertransform (FFT) circuit [9], [11], or downconvert a spectralslice of the signal to the electrical domain first using a coherentreceiver followed by a digital FFT [3], [8]. The second methodis preferable for long-haul systems because it enables DSPcompensation of impairments like chromatic dispersion (CD)before demultiplexing, thus avoiding the need for large guardbands. System complexity is also reduced by avoiding the needfor high-precision optical sampling and clock synchronization.

0733-8724/$31.00 2012 IEEE

HUANG et al.: HIGH-CAPACITY FIBER FIELD TRIAL 547

Fig. 1. Reported field trial results for capacity Tb/s (star: results achievedin our separate field trials; triangles: previously reported field trials, and dashedline: an equal value curve of capacity distance products).

In this paper, we report two separate field trials demon-strating the promising high capacities delivered by opticalsuperchannel transmission technology. We first describe theoperation principle of AO-OFDM and present the simulationresults for the purpose of estimating our trial performance inSection II. In Section III, we reported a record capacity of21.7 Tb/s with a mixed data rate channel plan over a long-hauldistance of more than 1500 km on field-installed standardsingle-mode fiber (SSMF) using AO-OFDM superchannels.Each subcarriers within the superchannels are modulated witheither dual-polarization 8 quadrature amplitude modulation(DP-8QAM) or DP-QPSK per subcarrier depending on theoptical signal-to-noise ratio (OSNR) after transmission [12],[13]. To extend system reach, in Section IV, we report 16.2 Tb/stransmission over 32 79.1 km spans of field-installed SSMFusing only erbium-doped fiber amplifiers (EDFAs) after eachspan [14]. Using superchannels with DP-QPSK zero-guard-in-terval subcarriers, we can increase the spectral efficiency by67% while retaining similar reach distance compare with 100Gb/s DP-QPSK systems [3]. Section V will summarize theresults of the two field trials.Fig. 1 shows our two field trial results compared with the

results of previously reported field trials. To the best of ourknowledge, both the capacity of 21.7 Tb/s trial and the ca-pacitydistance product (40.9 Pb/s km) of the 16.2 Tb/s trialare the highest reported to date in real-world field conditions.

II. AO-OFDM TRANSMISSION PRINCIPLE AND SIMULATION

A. Optical Superchannel Using AO-OFDMThe principle of generating AO-OFDM signal by parallel

subcarrier modulation is shown in Fig. 2. The key element thatdistinguishes AO-OFDM with a generic wavelength divisionmultiplexing (WDM) source is the multicarrier generator,whose output optical spectral has frequency components thatare equally spaced and phase synchronized. Several methodsexist for realizing such a multicarrier source, including theuse of a mode-locked laser synchronized at its fundamentalfrequency [9], the use of concatenated, overdriven opticalmodulators [15], or the use of a recirculating loop where a

new tone is generated at each pass through the loop [16].A tone demultiplexertypically an array waveguide grating(AWG)separates the frequency components of the opticalsource, which are then modulated with data via a parallelbank of optical modulators. The modulation method used ateach optical subcarrier is arbitrary, and can be SC with non-return-to-zero waveforms or digitally generated sinc-functionwaveforms (i.e., Nyquist pulse shaping); or OFDM signalsusing digitally generated waveforms. Different modulationformats, such as -ary PSK ( -PSK) or -ary quadratureamplitude modulation ( -QAM), can be selected for datatransmission based on system requirements such as distanceand spectral efficiency. In this study, we chose to use SCtransmission per optical subcarrier to avoid the requirementof DACs with large bandwidth, which are needed for Nyquistpulse shaping or electronic OFDM signal generation. In the SCcase, the baud rate per subcarrier ( where is thesymbol period) should be less than or equal to the frequencyspacing of the optical subcarriers to ensure that subcarrierdemultiplexing can be performed with low distortion.To demultiplex the subcarriers, a digital FFT can be imple-

mented easily in the receiver DSP alongside other digital func-tions already implemented in the current coherent 100G systemssuch as electronic CD compensation. Note that CD destroys or-thogonality between subcarriers. If an optical Fourier transformcircuit is used, CD must first be compensated precisely, whichcan be difficult to do using only optical CD compensators. Moreinformation on the operating principles, transmitter design, andassociated DSP algorithms for AO-OFDM superchannels canbe found in [15] and [17].

B. DP-8QAM Superchannel Design and SimulationTo achieve the initial goal of doubling existing field fiber

capacity in C-band while maintaining long-distance reach,AO-OFDM optical superchannel was selected for implementa-tion during the field trial to remove most of the frequency guardbands. DP-8QAMwas chosen as the desired modulation formatdue to its 50% higher spectral efficiency compared to DP-QPSKwith relative low OSNR requirement compared to DP-16QAM,making it still suitable for long-distance transmission. Eachof the DP-8QAM optical superchannel will contain a total of15 phase-locked optical subcarriers with 12.5 GHz spacing.With each subcarrier tributary carrying 75 Gb/s data, the rawtotal data rate and bandwidth of each superchannel are 1.125Tb/s and 200 GHz, which will result in a spectral efficiencyof 5.25 b/s/Hz after discounting the 7% overhead imposed byhard-decision forward error correction (HD-FEC) code.A hybrid amplification scheme using backward Raman

pumping and EDFA is targeted for the trial to obtain therequired OSNR for the DP-8QAM superchannels after trans-mission [12]. To estimate the transmission performance such assystem reach and to obtain system parameters such as optimumsignal launch power, we performed simulation for DP-8QAMsuperchannel transmission using am-plification scheme on VPI TransmissionMaker. The simulationsetup and conditions are plotted in Fig. 3. Since the simulationinvolves backward Raman pumping, which could be very timeconsuming if the simulated bandwidth is large, we limited the


Fig. 2. Principle of AO-OFDM using optical multicarrier technology.

Fig. 3. (a) Simulation setup for DP-8QAM optical superchannel transmission with backward Raman amplification. (b) Optical spectrum showing the wavelengthsof the Raman pumps and (c) signal power evolution curve (per channel) showing Raman onoff gain.

optical bandwidth to contain just two 1.125 Tb/s DP-8QAM op-tical superchannels, as shown in Fig. 3(a). The total transmis-sion bandwidth is 400 GHz, which is sufficient for cross-phasemodulation nonlinearity estimation under normal launch powerconditions. The two superchannels are separated by a guardband of 12.5 GHz, equivalent to the width of one subcarrier.For the fiber link, a concatenation of km spans ofSSMF with adjusted loss of 0.25 dB/km was used to simulatethe higher loss per span measured on the field-installed fibers.The Raman pump parameters were obtained by measuring thepumping output of an actual three-pump Raman amplifier witha total pump power of 27 dBm, which would be used in thetrial. The optical spectrum of the Raman pumps is plotted inFig. 3(b). Fig. 3(c) plots the signal power evolving over the dis-tance of one fiber span, showing a 9.15 dB Raman onoff gainwhen the launch power per superchannel is set at 3 dBm. Thelink EDFA noise figure is set at 4.5 dB in the simulation. An op-tical spectral power equalizer was inserted every five spans toequalize the signal gain tilting, as shown in Fig. 3(a) inset. A dig-ital coherent receiver, implemented in MATLAB, then down-converted the subcarrier of interest to baseband and use DSP

Fig. 4. Simulated signal Q versus distance usingunder different signal launch power.

to compensate the fiber link impairments, demultiplex the sub-carrier in two orthogonal polarizations, and restore the 8-QAMsignal constellation.


Fig. 5. Experimental setup for the high-capacity field trial; solid lineoptical signals, dotted lineelectrical signals.

Fig. 6. Optical spectra captured after (a) wideband phase modulation to generate phase-locked optical tones for odd and even superchannel groups. (b) Beforeand after passive recombination of modulated odd and even optical subcarrier groups. (c) Spectrum of 22 optical superchannels for transmission.

Fig. 4 shows the received signal Q-factor (defined as re-ceived signal-to-noise ratio per symbol) versus transmissiondistance (up to 20 spans) using different launch powers fortwo 1.125 Tb/s superchannels. Back-to-back (BTB) noise wasinserted to cap the BTB Q-factor at 20 dB. Assuming the lossis homogeneous across the whole fiber span, the 8 dBm launchpower curve achieved the highest Q-factor for distance longerthan 15 spans, equivalent to 5 dBm per superchannel. Basedon HD-FEC coding, the Q-factor requirement for 8-QAM is12.5 dB. Therefore, our simulation results show that 1600 kmtransmission is achievable for DP-8QAM with around 2 dBmargin. The nonsmooth Q-factor transitions at certain shorterspan distances are mainly due to the short symbol sequencewe used in order to save simulation time, which could causereceiver DSP to converge at different conditions. It is best tonote that this result is based on two superchannel transmissionwith the assumption of flat EDFA gain and low noise figure.With signal populated over the entire C-band, as in the targetedtrial, the received signal OSNR will drop due to uneven gainand noise figure distribution, therefore reducing the eventualperformance.

III. 21.7 TB/S FIELD TRIAL USING HYBRIDDP-8QAM/DP-QPSK-MODULATED OPTICAL SUPERCHANNELS

A. Experimental Configuration of the Field Trial

The experimental setup is shown in Fig. 5. A total of 22 ex-ternal cavity lasers (ECLs), centered from 191.75 to 195.9375THz are used as seed lasers and combined into odd

and even groups using two AWGs (100 GHz).Phase-locked optical carriers at 25 GHz spacing are generatedfor each seed laser by using separate phase modulators over-driven with sine waves [3]. To avoid crosstalk during carriergeneration, separate phase modulators are used for odd and evenlasers [respected outputs are shown in Fig. 6(a)]. A flexible-band wavelength selective switch (WSS) simultaneously equal-izes the powers and adjusts the phases of the unmodulated op-tical subcarriers, filters out undesired subcarriers, and combinesthe subcarriers from the two groups into one output [18], [19].A novel flexible format modulation unit is used to create

8QAM or QPSK modulation [12] at 12.5 Gbaud using three bi-nary drive signals with pseudorandom binary sequence (PRBS)


length of . The modulation unit consists of a Carte-sian I/Q modulator, which has a structure of parallel nestedMachZehnder modulators (MZMs) to produce a QPSK output,followed by a dual-drive MZM. The time delay between thetwo binary drive signals to the I/Q modulator is symbols.For QPSK operation, the drive signal to the dual-drive MZMis turned off. For 8QAM generation, the QPSK signal from I/Qmodulator is passed on to the dual-drive MZM. One port of thedual-drive MZM is grounded, while the other port is driven bya third binary signal which is the inversion of the other two andhas a time delay difference of symbols. The drive signalwill synchronously modulate the amplitude of the QPSK inputsignal and simultaneously toggle the output phase by inorder to map the symbol to either the inner or the outer rings inthe 8QAM constellation, as shown in Fig. 5 inset. The requireddriving voltage swing and the bias voltage can bedetermined by solving the following equations, with the MZMoutput amplitude and phase :

(1)

(2)

It is found that with bias set at off theMZM quadrature point. This flexible transmitter design allowsus to dynamically select the subcarrier modulation format basedon the spectral efficiency requirement and OSNR condition.The modulated subcarriers, denoted as odd subcarriers, are

then split into two paths. An I/Q modulator with sine and cosinewaves driving its two arms is employed to shift the spectrum ofthe signals in one path by GHz to generate even sub-carriers. In another path, the odd subcarriers are decorrelated intime by symbols before they are passively combined withthe even subcarriers with matched symbol timing. Both odd andeven subcarriers are shown in Fig. 6(b). An automatic feedbackloop is implemented for subcarrier power equalization. An op-tical spectral analyzer first captures the spectrum of the odd sub-carriers. The spectrumwas read continuously by a computer andused to generate the corresponding attenuation values to updatethe WSS profile. With this automatic control loop, the intensityfluctuation of each individual subcarriers can be minimized tomaintain the flatness of the signal spectrum. Polarization multi-plexing is done by splitting, delaying one copy of the signals by

symbols, followed by orthogonal polarization recombina-tion. The transmitter output spectrum is plotted in Fig. 6(c), as atotal of 22 optical superchannels are generated and multiplexedin wavelength with 12.5 GHz guard band. Each of the odd super-channels contains sixteen 12.5 Gbaud subcarriers, while each ofthe even superchannels contains fourteen. The total bandwidthof the WDM superchannels is 4.4 THz (35.2 nm) including theguard bands, containing 18 DP-8QAM superchannels (9 1.2and 9 1.05 Tb/s) and four DP-QPSK superchannels (2 0.8and 2 0.7 Tb/s).

The WDM superchannels were transmitted over a link con-sisting of 19 dispersion-uncompensated fiber spans with a totaldistance of 1503 km. Each span contains a 79.1 km installedfield fiber, which is a part of Verizons long-haul network. Thefiber link was dedicated only to our trial when the experimentwas performed. The average fiber loss is 20.8 dB per span,which is compensated by a hybrid scheme of backward Ramanpumping (27 dBm average pump power) with EDFA. The aver-aged CD and slope of the fibers are 16.75 ps/nm/km and 0.055ps/nm /km at 1550 nm, respectively, and the averaged meanDGD of the fiber spans is 0.65 ps/ km. From our simulationresult in the previous section, the optimum launch power persuperchannel is 5 dBm. Therefore, the total launch power for22 superchannels at beginning of each span is set atdBm. A WSS with 50 GHz resolution is inserted every threespans to compensate the gain tilt caused by the EDFAs andRaman pumps. At the receiver, another WSS was used to filterout a single superchannel each time for measurement. The fil-tered superchannel was then sent to the input of a polariza-tion-diverse 90 optical hybrid, where the spectral slice of in-terest was downconverted to baseband by tuning the optical fre-quency of the local oscillator (LO) laser. The outputs of the op-tical hybrid were detected using balanced photodetectors with40 GHz bandwidth, recovering the in-phase (I) and quadrature(Q) signals in the X- and Y-polarization of the selected spec-tral slice. These signals are sampled and digitized using twoLeCroy real-time sampling oscilloscopes with 80 GS/s sam-pling rate and a single-side electrical bandwidth of 30 GHz. Thefour scope channels have effective number of bits ranging fromfive bits at low frequency to four bits at high frequency. TheLO was tuned to midway between two optical carriers to detecttwo subcarriers per data capturing. Five datasets of 2 s dura-tion were captured and processed. To detect the subcarrier of in-terest in DSP, we recentered to DC by multiplying the capturedsignal with a numerical LO at GHz before we processeach separately using an offline DSP which includes a fixedfrequency-domain equalizer (FDE) for dispersion compensa-tion, followed by an adaptive time-domain equalizer (TDE) tocompensate nonstationary effects. To suppress demultiplexingcrosstalk, the FDE and TDE operate at 6 oversampling (75GSa/s). Within the TDE, constant-modulus algorithm (CMA) isfirst used and then followed by decision-aided adaptation to ini-tialize the equalizer using training sequences before the processis transferred to decision-directed (DD) adaptation mode [20].After convergence of the TDE, the second half of each datasetis used for bit error rate (BER) counting (62 500 symbols persubcarrier).

B. Results and DiscussionsFig. 7(a) depicts the optical spectrum for all 22 optical

superchannels after transmission and Fig. 7(b) plots the mea-sured BERs for the field trial. For the first 18 superchannels

the measured OSNR (measured at 0.1 nm reso-lution for both signal and noise) after 1503 km transmissionis between 15.5 and 16.5 dB, yielding signal Q-factors largerthan 12.5 dB using DP-8QAM modulation. For the 18 su-perchannels, the averaged spectral efficiency is 5.26 b/s/Hzafter discounting a 7% of HD-FEC overhead. The last four


Fig. 7. (a) Received optical spectrum of 22 optical superchannels. (b) Measured BERs for all 330 optical subcarriers with inset showing constellations for hybridDP-8QAM and DP-QPSK modulations.

superchannels on the short wavelength side haslower received OSNR between 13.515 dB which is mainlydue to lower gain and higher amplified spontaneous emissionnoise from the EDFAs. The low OSNR values are not sufficientto support DP-8QAM superchannels; therefore, the modulationformat was switched to DP-QPSK using the flexible mod-ulation unit. The averaged spectral efficiency in this regionhas been reduced to 3.5 b/s/Hz. With this configuration, theBER values of all 330 optical subcarriers in 22 superchannelsare below the threshold of 4.5 10 using a product-readycontinuously interleaved BoseChaudhuriHocquenghem FEC[21], which corresponds to dB Q-factor. Compared toour simulation in the previous section, the drop in trial per-formance is contributed by lower B2B Q-factor, nonperfectlygenerated 8-QAM constellation, and mainly the higher EDFAnoise figure when operated in full C-band with gain imbalanceacross the transmission bandwidth. The lower BER values for

is due to the fact that DP-QPSK modulation haslarger transmission margin than its DP-8QAM counterparts.The performance fluctuations of the subcarriers are mainlycaused by the intensity difference between neighboring sub-carriers and also the OSNR distribution. Having the ability toadminister different modulation formats based on OSNR per-formance allows us to use a larger portion of the spectrum withadjustable spectral efficiency, to ensure signal performance,and to obtain the highest total capacity possible at 21.7 Tb/safter discounting for 7% FEC overhead.

IV. 16.2 TB/S FIELD TRIAL USING DP-QPSK-MODULATEDOPTICAL SUPERCHANNELS

A. Experimental Configuration of the Field TrialThe trial results in the previous section have shown that the

DP-8QAM superchannel system was limited in transmissiondistance by the higher OSNR requirement. A second trial wasconducted using DP-QPSK-modulated AO-OFDM superchan-nels to significantly increase the system reach while sacrificingsystem capacity as a tradeoff. The experimental setup, shown inFig. 8(a), is similar to the previous section with minor changes

to accommodate for differences such as modulation formatand optical bandwidth per superchannel. Sixteen ECLs withcenter wavelengths ranging from 191.6375 to 196.15 THz areused. Odd- and even- wavelengths areseparately combined using two 100 GHz AWGs. For opticalmultitone generation, similar technique as shown in Section IIIis employed, where wideband phase modulation generatephase-locked carriers at 25 GHz spacing separately for odd andeven lasers. After equalization and combination of the opticalcarriers by the flexible-band WSS, we split the signal into twopaths. In one of the paths, an I/Q modulator configured as asingle-sideband modulator generates optical carriers at 12.5GHz offset from the original signal, such that after recombining,each superchannel contains unmodulated optical comb spacedat 12.5 GHz. For data modulation, we first use a 12.5/25 GHzinterleaver to divide the subcarriers into even and odd groups.Each group is separately modulated with QPSK symbols at 12.5Gbaud chosen from a PRBS of length . The modulatedsubcarriers are passively combined with the correct symbolalignment [spectrum shown in Fig. 8(b)]. Finally, polarizationmultiplexing is performed by splitting the signal, delaying onecopy by symbols, followed by orthogonal polarizationrecombination. The 16 optical superchannels are multiplexedin wavelength with 12.5 GHz guard band. Each of the oddsuperchannels has 24 subcarriers (1.2 Tb/s), while each of theeven superchannels have 22 subcarriers (1.1 Tb/s). The totalWDM signal bandwidth is 4.8 THz (38.4 nm).The WDM superchannels were transmitted over an installed

field fiber link consisting of 32 79.1 km dispersion-uncom-pensated fiber spans. The average loss, CD value and slope, andmean DGD remain the same as the measured values stated inthe previous section. The signal is amplified by an EDFA onlyafter each span. We adjusted the launch power to be 20 dBm atthe beginning of every span. For gain tilt compensation, a WSSwith 50 GHz resolution is inserted every three spans. A digitalcoherent receiver, same as the one in the previous trial, is usedto detect two optical subcarriers for each data capture by tuningthe LO midway between their center frequencies. Five datasetsof 2 s duration were captured and processed using an offline


Fig. 8. (a) Experimental setup for the field trial using DP-QPSK optical superchannels. (b) Optical spectra captured after odd/even subcarrier modulation andrecombination. (c) Transmitted optical spectra for the 16 WDM DP-QPSK optical superchannels.

Fig. 9. (a) Received optical spectra for the 16 WDM DP-QPSK optical superchannels after transmission. (b) Measured BER results for all 346 subcarriers.

DSP which includes a fixed FDE for dispersion compensation,followed by an adaptive TDE to compensate nonstationary ef-fects. For QPSK modulation, 4 oversampling (50 GSa/s) issufficient to suppress the demultiplexing crosstalk between thesubcarrier in the FDE and TDE. The TDE is first adapted usingCMA, followed by DD adaptation [3]. After convergence of theTDE, the second half of each dataset is used for BER counting(62 500 symbols per subcarrier).

B. Results and DiscussionsThe WDM superchannel spectra before and after transmis-

sion are plotted in Figs. 8(c) and 9(a). About half of the firstsuperchannel on the longer wavelength side is located inthe gain roll-off region of the EDFAs. We therefore pre-em-phasize the subcarriers using a flexible-band WSS before trans-mission, so only three subcarriers are dropped due to the gainroll-off. On the shorter wavelength side, due to the bandwidthmismatch of the link WSSs purchased from different vendors,

all but four subcarriers from the last superchannels areblocked during transmission. The OSNRs for the remaining 346subcarrier were between 11 and 12 dB (measured at 0.1 nm res-olution for both signal and noise) after 2531 km. Fig. 9(b) showsthe measured BERs for all 346 subcarriers. All of the BERs arebelow the HD-FEC threshold. The fluctuation in BER is mainlycaused by difference in power between neighboring subcarriers.After accounting for FEC overhead, the achieved spectral effi-ciency is 3.58 b/s/Hz. Comparing with our previous field trialusing DP-8QAM modulation [13], the transmission distance isincreased by 68%with only a 25% decrease in capacity, yieldinga total capacitydistance product of 40.9 Pb/s km.

V. CONCLUSIONWe have successfully performed two separate field trials

achieving the highest fiber capacity and highest capacitydis-tance product reported to date. In the record high-capacity 21.7Tb/s field fiber transmission experiment, we used a total of 22


optical superchannels with a flexible band WDM and 1503 kmof field installed fiber employing EDFAs and Raman amplifiers.All channels are multisubcarrier superchannels generated usingthe AO-OFDM technique with hybrid DP-8QAM/DP-QPSKmodulation at each subcarrier. A novel dynamic modulationformat selector is also used to combat uneven OSNR distribu-tion within the transmitted spectrum. A coherent receiver wasused to select the subcarriers of interest by tuning the LO fre-quency. To extend system reach distance, we then transmittedoptical superchannels with DP-QPSK modulated subcarriersover 2531 km of field-installed fiber. A net capacity of 16.2Tb/s was transmitted using 16 optical superchannels on flex-ible-grid WDM with EDFA-only amplification. The obtained40.9 Pb/s km capacitydistance product is, to the best of ourknowledge, the highest reported to date over field-installedfiber. The two separate trials demonstrated that fiber capacitycan be significantly increased from current 100 Gb/s systemsusing multicarrier optical superchannels. Our trial results alsoproved that it is feasible to double the current capacity on aninstalled standard fiber infrastructure at long-haul distances.

REFERENCES[1] Verizon deploys commercial 100G ultra-long-haul optical system on

portion of its core European network Dec. 14, 2009, Verizon Press Re-lease.

[2] T. Kobayashi, S. Yamanaka, H. Kawakami, S. Yamamoto, A. Sano,H. Kubota, A. Matsuura, E. Yamazaki, M. Ishikawa, K. Ishihara, T.Sakano, E. Yoshide, Y. Miyamoto, M. Tomizawa, and S. Matsuoka,8-Tb/s (80 127Gb/s) DP-QPSKL-bandDWDMTransmission over457-km Installed DSF Links with EDFA-only amplification, in Proc.15th OptoElectron. Commun. Conf., Sapporo, Japan, Jul. 2010, pp.12.

[3] T. J. Xia, G. A. Wellbrock, Y.-K. Huang, E. Ip, M.-F. Huang, Y. Shao,T.Wang, Y. Aono, T. Tajima, S. Murakami, andM. Cvijetic, Field ex-periment with mixed line-rate transmission (112-Gb/s, 450-Gb/s, and1.15-Tb/s) over 3,560 km of installed fiber using filterless coherent re-ceiver and EDFAs only, in Proc. Opt. Fiber Commun. Conf. Expo./Nat. Fiber Opt. Eng. Conf., Los Angeles, CA, Mar. 2011, pp. 13.

[4] D. Qian, M.-F. Huang, E. Ip, Y.-K. Huang, Y. Shao, J. Hu, andT. Wang, 101.7-Tb/s (370 294-Gb/s) PDM-128QAM-OFDMtransmission over 3 55-km SSMF using Pilot-based phase noisemitigation, in Proc. Opt. Fiber Commun. Conf. Expo./Nat. Fiber Opt.Eng. Conf., Los Angeles, CA, Mar. 2011, pp. 13.

[5] A. Sano, T. Kobayashi, S. Yamanaka, A. Matsuura, H. Kawakami, Y.Miyamoto, K. Ishihara, and H. Masuda, 102.3-Tb/s (224 548-Gb/s)C- and extended L-band all-Raman transmission over 240 km usingPDM-64QAM single carrier FDMwith digital pilot tone, inProc. Opt.Fiber Commun. Conf. Expo./Nat. Fiber Opt. Eng. Conf., Los Angeles,CA, Mar. 2012, pp. 13.

[6] X. Yi, N. K. Fontaine, R. P. Scott, and S. J. B. Yoo, Tb/s coherent op-tical OFDM systems enabled by optical frequency combs, J. Lightw.Technol., vol. 28, no. 14, pp. 20542061, Jul. 2010.

[7] Y. Ma, Q. Yang, Y. Tang, S. Chen, and W. Shieh, 1-Tb/s single-channel coherent optical OFDM transmission over 600-km SSMF fiberwith subwavelength bandwidth access, Opt. Exp., vol. 17, no. 11, pp.94219427, May 2009.

[8] S. Chandrasekhar, X. Liu, B. Zhu, and D. W. Peckham, Transmis-sion of a 1.2-Tb/s 24-carrier no-guard-interval coherent OFDM su-perchannel over 7200-km of ultra-large-area fiber, in Proc. 35th Eur.Conf. Opt. Commun., Vienna, Austria, Sep. 2009, pp. 12.

[9] D. Hillerkuss, T. Schellinger, R. Schmogrow, M. Winter, T. Vallaitis,R. Bonk, A. Marculescu, J. Li, M. Dreschmann, J. Meyer, S. B. Ezra,N. Narkiss, B. Nebendahl, F. Parmigiani, P. Petropoulos, B. Resan, K.Weingarten, T. Ellermeyer, J. Lutz, M. Mller, M. Huebner, J. Becker,C. Koos, W. Freude, and J. Leuthold, Single source optical OFDMtransmitter and optical FFT receiver demonstrated at line rates of 5.4and 10.8 Tbit/s, in Proc. Opt. Fiber Commun. Conf. Expo./Nat. FiberOpt. Eng. Conf., San Diego, CA, Mar. 2010, pp. 13.

[10] Y.-K. Huang, D. Qian, J. Yu, and T. Wang, 150 Gb/s PolMUX-8PSKall-optical OFDM using digital coherent detection with modified CMAalgorithm, in Proc. LEOS Annu. Meet. Conf., Antalya, Turkey, Oct.2009, pp. 4647.

[11] A. J. Lowery, Design of arrayed-waveguide grating routers for useas optical OFDM demultiplexers, Opt. Exp., vol. 18, no. 13, pp.1412914143, Jun. 2010.

[12] Y.-K. Huang, E. Ip, P. N. Ji, Y. Shao, T. Wang, Y. Aono, Y. Yano,and T. Tajima, Terabit/s optical superchannel with flexible modula-tion format for dynamic distance/route transmission, presented at thepresented at the Opt. Fiber Commun. Conf. Expo./Nat. Fiber Opt. Eng.Conf., Los Angeles, CA, Mar. 2012, Paper OM3H.4.

[13] T. J. Xia, G. A. Wellbrock, Y.-K. Huang, M.-F. Huang, E. Ip, P. Ji, D.Qian, A. Tanaka, Y. Shao, T. Wang, Y. Aono, and T. Tajima, 21.7Tb/s field trial with 22 DP-8QAM/QPSK optical superchannels over1,503-km of installed SSMF, in Proc. Opt. Fiber Commun. Conf.Expo./Nat. Fiber Opt. Eng. Conf., Los Angeles, CA, Mar. 2012, pp.13.

[14] Y.-K. Huang, M.-F. Huang, E. Ip, E. Mateo, A. Tanaka, D. Qian, Y.Shao, T. Wang, Y. Aono, T. Tajima, T. J. Xia, and G. A. Wellbrock,16.2-Tb/s field trial over 2,531-km of installed SSMF with DP-QPSKoptical superchannels, in Proc. 17th OptoElectron. Commun. Conf.,2012, pp. 3334.

[15] Y.-K. Huang, E. Ip, Z. Wang, M.-F. Huang, Y. Shao, and T. Wang,Transmission of spectral efficient super-channels using all-op-tical OFDM and digital coherent receiver technologies, J. Lightw.Technol., vol. 29, no. 24, pp. 38383844, Dec. 2011.

[16] J. Yu, Z. Dong, X. Xiao, Y. Xia, S. Shi, C. Ge, W. Zhou, N. Chi, andY. Shao, Generation, transmission and coherent detection of 11.2 Tb/s(112 100 Gb/s) single source optical OFDM superchannel, in Proc.Opt. Fiber Commun. Conf. Expo./Nat. Fiber Opt. Eng. Conf., Los An-geles, CA, Mar. 2011, pp. 13.

[17] Y.-K. Huang, E. Ip, T. J. Xia, G. A. Wellbrock, M.-F. Huang, Y. Aono,T. Tajima, and M. Cvijetic, Mixed line-rate transmission (112-Gb/s,450-Gb/s, and 1.15-Tb/s) over 3560 km of field-installed fiber withfilterless coherent receiver, J. Lightw. Technol., vol. 30, no. 4, pp.609617, Feb. 2012.

[18] G. Baxter, S. Frisken, D. Abakoumov, H. Zhou, I. Clarke, A. Bartos,and S. Poole, Highly programmable wavelength selective switchbased on liquid crystal on silicon switching elements, presented atthe presented at the Opt. Fiber Commun. Conf. Expo./Nat. Fiber Opt.Eng. Conf., Anaheim, CA, Mar. 2006, Paper OTuF2.

[19] M. A. Roelens, J. A. Bolger, D. Williams, and B. J. Eggleton, Multi-wavelength synchronous pulse burst generation with a wavelength se-lective switch, Opt. Exp., vol. 16, pp. 1015210157, Jul. 2008.

[20] E. Ip, N. Bai, Y.-K. Huang, E. Mateo, F. Yaman, M.-J. Li, S. Bickham,S. Ten, J. Linares, C. Montero, V. Montero, X. Prieto, V. Tse, K. M.Chung, A. Lau, H.-Y. Tam, C. Lu, Y. Luo, G.-D. Peng, and G. Li,88 3 112-Gb/s WDM transmission over 50 km of three-modefiber with inline few-mode fiber amplifier, in Proc. 37th Eur. Conf.Opt. Commun., Geneva, Switzerland, Sep. 2011, pp. 13.

[21] M. Scholten, T. Coe, and J. Dillard, Continuously-interleaved BCH(CI-BCH) FEC delivers best-in-class NECG for 40G and 100G metroapplications, in Proc. Opt. Fiber Commun. Conf. Expo./Nat. FiberOpt. Eng. Conf., San Diego, CA, Mar. 2010, pp. 13.

Author biographies not included by author request due to spaceconstraints.

high-capacity fiber field trial

Documents