a broadcast-capable wdm passive optical network using offset polarization multiplexing

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 14, JULY 15, 2012 2329

A Broadcast-Capable WDM Passive Optical NetworkUsing Offset Polarization Multiplexing

Fei Xiong, Student Member, IEEE, Wen-De Zhong, Senior Member, IEEE, and Hoon Kim, Senior Member, IEEE

Abstract—We propose and demonstrate a wavelength-divi-sion-multiplexed passive optical network providing both unicastand broadcast services using an offset polarization multiplexingtechnique. The downstream differential phase-shift keying(DPSK)-formatted unicast and broadcast signals are offset polar-ization-multiplexed at the central office and demultiplexed anddetected at the optical network units without resorting to anypolarization tracking. A portion of the offset polarization-mul-tiplexed downstream signals is fed into a polarization-sensitiveweak-resonant-cavity Fabry–Perot laser diode for upstreamremodulation without any polarization control. Successful trans-missions of 10-Gb/s downstream unicast and broadcast DPSKsignals as well as 2.5-Gb/s upstream ON–OFF keying signalover a 20-km standard single-mode fiber are experimentallydemonstrated. The robustness of the proposed scheme againstpolarization fluctuation along the link, relative bit delay betweenthe unicast and broadcast signals, frequency deviation of thedownstream signals from the delay interferometer (DI), andimperfection of the DI is investigated.

Index Terms—Broadcast, Fabry–Perot laser diode (FP-LD), po-larization multiplexing, wavelength-division-multiplexed passiveoptical network (WDM-PON).

I. INTRODUCTION

W AVELENGTH-DIVISION-MULTIPLEXED PAS-SIVE OPTICAL NETWORK (WDM-PON) is

considered as an attractive solution for the next-generationbroadband access network. In addition to its large guaran-teed bandwidth, WDM-PON offers other advantages suchas high security, flexibility, and upgradability. However, thelogical point-to-point topology of WDM-PON hinders directand simple delivery of broadcast/multicast services. In orderto meet the ever-increasing demand of broadcast/multicastservices, e.g., video-on-demand and high-definition televisiondistribution services, several approaches have been proposed todeliver broadcast or multicast services over WDM-PON archi-tectures [1]–[7]. One straightforward method was to allocateadditional wavelength channels for broadcast data delivery

Manuscript received January 17, 2012; revised April 13, 2012; accepted April19, 2012. Date of publication May 03, 2012; date of current version June 06,2012.F. Xiong and W.-D. Zhong are with the School of Electrical and Electronic

Engineering, Nanyang Technological University, Nanyang Avenue, Singapore637553 (e-mail: [email protected]; [email protected]).H. Kim is with the Department of Electrical and Computer Engineering, Na-

tional University of Singapore, Singapore 117576 (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.2012.2196756

[1], [2]. However, this method induced low spectral efficiency[1] or required high-power broadcast signal to compensatethe large optical loss caused by a power splitter at the remotenode (RN) [2]. In [3], the subcarrier multiplexing (SCM)technique was adopted to accommodate multicast services.However, high-speed optoelectronic components were neededin both optical line terminals (OLTs) and optical network units(ONUs). Another approach was to exploit two orthogonalmodulation formats [e.g., ON–OFF keying (OOK) and dif-ferential phase-shift keying (DPSK)], one for unicast servicesand the other for multicast services [4]. The extinction ratio(ER) of the unicast OOK signal had to be sacrificed to enablethe multicast services. In [5], the broadcast DPSK signal wastime-interleaved with the unicast DPSK signal, in which timingcontrol was needed in both OLTs and ONUs.Recently, we have proposed and experimentally demon-

strated a WDM-PON architecture providing broadcast overlayby exploiting polarization multiplexing [6], [7]. In thisWDM-PON architecture, the downlink unicast and broadcastdata are, respectively, carried by two orthogonally polar-ized optical beams from a single light source. Not only doesthis technique support broadcast services without allocatingadditional wavelength channels and using high-frequencySCM modulation, but it also depolarizes the seeding light toFabry–Perot laser diodes (FP-LDs) for upstream transmissionand thus can be applied to low-cost polarization-sensitive color-less upstream optical transmitters. However, active polarizationtracking is required to demultiplex the two polarization-mul-tiplexed signals at each ONU, which may hinder the realdeployment of such a system.To eliminate the need for active polarization tracking, we

here propose and demonstrate a new WDM-PON architecturein which offset polarization multiplexing is exploited to sup-port broadcast capability. The DPSK-formatted downstreamunicast and broadcast signals whose wavelengths are slightlydifferent from each other are combined through polarizationmultiplexing. At each ONU, the downstream DPSK-formattedunicast and broadcast signals are demultiplexed and demodu-lated by two DPSK demodulators without any need of activepolarization tracking. The scheme also completely depolarizesthe seeding light for the upstream FP-LDs, which enables con-stant injection-locking of the polarization-sensitive FP-LDs.Simultaneous transmissions of 10-Gb/s downstream unicastand broadcast DPSK signals as well as 2.5-Gb/s upstreamOOK signal are experimentally demonstrated. The effects ofRayleigh backscattering (RB), remodulation crosstalk, polar-ization fluctuation along the link, relative bit delay betweenthe unicast and broadcast signals, frequency deviation of the

0733-8724/$31.00 © 2012 IEEE

2330 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 14, JULY 15, 2012

Fig. 1. Operation principle of demultiplexing/demodulation of the offset polarization-multiplexed DPSK signals. In this figure, the x-polarized signal is to bedetected, and thus, the y-polarized signal becomes crosstalk to the x-polarized signal. To detect the y-polarized signal, we need another DI whose peak or null isaligned to the center wavelength of the y-polarized signal.

downstream signals from the delay interferometer (DI), andimperfection of the DI on the transmission performances of thedownstream and upstream signals are investigated in detail byboth experiments and simulations.

II. OPERATION PRINCIPLE OF THE PROPOSEDWDM-PON ARCHITECTURE

In the offset polarization multiplexing scheme, twoDPSK-formatted optical signals separated by a frequencyoffset are polarization-multiplexed at the transmitter.At the receiver, they are demultiplexed and demodulated bytwo DIs preceded by an optical coupler, without using anyactive polarization tracking and polarization beam splitters[8]. In DPSK reception, a sensitivity penalty arises when thelaser frequency deviates from the DI frequency. The frequencydeviation between the laser frequency and the DI is relatedto phase deviation through , whereis the time delay within the DI [9]. Penalty-free detectionof a DPSK signal is achieved when the phase deviation isan integer multiple of , whereas phase deviation of an oddmultiple of completely closes eye opening of the detectedsignal, leading to infinite power penalty, as depicted in Fig. 1.Therefore, if two offset polarization-multiplexed optical signalsare separated by odd multiples of in frequency (i.e.,

, where is an integer) and one ofthe optical signals is aligned to a DI having a free spectral range(FSR) of such that the phase deviation is , the otheroptical signal would experience a complete eye closure. Asillustrated in Fig. 1, the optical signal with phase deviation ofzero (represented by the solid arrow) can be detected with min-imum crosstalk from the other optical signal locatedaway (represented by the dot–dashed arrow). It is noted thatpolarization multiplexing of the two optical signals helps tomake the crosstalk added in power rather than in E-field. Thissubstantially reduces the deleterious effects of crosstalk. For thedetection of the other optical signal, we need another DI whichis aligned to this optical signal with a zero phase deviation.The schematic of the proposed WDM-PON architecture

using offset polarization multiplexing is shown in Fig. 2. Inthe OLT, two distributed feedback (DFB) lasers separated

by odd multiples of in frequency are assigned tounicast and broadcast services, respectively, for each ONU.The unicast and broadcast data are modulated onto the cor-responding optical carriers in DPSK format through separateMach–Zehnder modulators (MZMs). The downstream unicastsignals are combined by an arrayed waveguide grating (AWG)and then polarization-multiplexed with the broadcast signalsusing a polarization beam combiner (PBC). The offset polar-ization-multiplexed signals are transmitted through a feederfiber to the RN where they are wavelength-demultiplexed byan AWG. At each ONU, a portion of the downstream signalsis divided into two equal parts for respective downstreamunicast and broadcast detections. The other portion is fed intoan FP-LD for upstream remodulation. The upstream signalis wavelength-multiplexed and transmitted back to the OLTthrough the same feeder fiber and detected at the OLT. In theproposed broadcast-enabled WDM-PON architecture shown inFig. 2, two sets of DFB lasers separated by are employedfor unicast and broadcast deliveries, respectively. An alternativeapproach, which can reduce the complexity and implementationcost, is to simultaneously generate the broadcast carriers byfrequency-shifting a portion of each of the unicast carriers byan amount of through subcarrier modulation using oneoptical single-sideband modulator (OSSBM) [10]. As shownin Fig. 3, the OSSBM is driven by an electrical sinusoidalsignal with a frequency of , resulting in single-sidebandcarrier-suppressed signals whose frequencies differ from theirrespective original carriers by an amount of . Theseresultant single-sideband carrier-suppressed signals servingas the broadcast carriers are then fed into another MZM andmodulated with broadcast data in DPSK format. In this way,only one set of DFB lasers are required in the OLT.Two DIs in each ONU would increase the cost at customer

premises. However, the additional cost incurred by two DIs ineach ONU could be justified by the increased capacity broughtby the proposed scheme. It doubles the spectral efficiencywithout doubling the line rate, or employing complicatedhigher order modulation formats, or using active polarizationtracking for polarization demultiplexing. It accommodatesmore end users than the scheme which assigns additional

XIONG et al.: BROADCAST-CAPABLE WDM PASSIVE OPTICAL NETWORK 2331

Fig. 2. Schematic of the proposed WDM-PON architecture using offset polarization multiplexing.

Fig. 3. Alternative structure of the OLT for the proposed WDM-PON scheme.

wavelengths to deliver broadcast services. It eliminates theperformance tradeoff between the downstream and upstreamsignals and the need of bit synchronization between the uni-cast and multicast data existed in the systems employing twoorthogonal modulation formats. Moreover, the DPSK receiverscould be implemented in compact and cost-effective manners,e.g., using silicon semiconductor manufacturing technology[11], [12].

III. EXPERIMENT SETUP AND RESULTS

Fig. 4 depicts the experimental setup to demonstrate theproposed scheme and to characterize the transmission perfor-mances. A DFB laser (LD1) and a tunable laser (LD2) wereseparately modulated through two MZMs. The frequency offset

between the two lasers was set to be ,where ps in the experiment, by tuning the wave-length of the tunable laser. The modulators were biased attheir transmission nulls for phase modulation and driven bytwo 10-Gb/s nonreturn-to-zero (NRZ) decorrelatedpseudorandom binary sequences (PRBSs) representing thedownstream unicast and broadcast data, respectively. A dif-ferential encoder was not employed since differentially codedPRBS is simply a time-delayed replica of the PRBS [9]. A

tunable optical delay line (ODL) was added to one of the signalpaths before the PBC to study the effect of the relative bit delaybetween the two downstream data on the system performance.The powers of the two signals were equalized at the output ofthe PBC by an optical attenuator (OA). It would balance thebit error rate (BER) performance of the downstream unicastand broadcast signals and also help to minimize the degreeof polarization (DOP) of the offset polarization-multiplexeddownstream signals to facilitate the external seeding of thepolarization-sensitive FP-LD used for upstream transmission.The offset polarization-multiplexed downstream signals

were transmitted through a 20-km standard single-mode fiber(SSMF). At the ONU, the downstream signals were divided intotwo portions by a 50/50 optical coupler, one for downstreamdetection and the other for upstream remodulation. The DPSKreceiver was composed of a DI with a 10-GHz FSR and a10-Gb/s avalanche photodetector (APD) single-ended receiver.A single DPSK receiver was used to detect either the unicastor broadcast data. In real implementation, two DPSK receiversare needed for each ONU to detect the unicast and broadcastsignals simultaneously. The other portion of the downstreamsignal was fed into a polarization-sensitive FP-LD which wasdirectly modulated by a 2.5-Gb/s NRZ PRBS with a pattern


Fig. 4. Experimental setup in (I) single-feeder-fiber bidirectional configuration and (II) two-feeder-fiber unidirectional configuration (only used for investigatingthe RB-induced power penalty).

length of . The FP-LD was an uncooled device housed ina transistor-outlook-can package. The seeding power wasdBm and there was no polarization controller (PC) insertedbetween the optical coupler and the FP-LD. After injection, theupstream signal was transmitted back to the OLT through thesame feeder fiber and detected by a 2.5-Gb/s APD receiver. Anoptical bandpass filter (BPF) having a 3-dB bandwidth of 40GHz was utilized before the upstream receiver to emulate anAWG at the OLT.In our single-ended detection, we utilized the destructive port

of the DI. This is because the return-to-zero-like waveform fromthe destructive port exhibits better receiver sensitivity than thesignal from the constructive port [13]. Fig. 5 shows the BERperformance of the downstream unicast and broadcast signals.When the broadcast service was disabled (i.e., LD2 was turnedOFF), the receiver sensitivity of the downstream unicast signalwas measured to be dBm at a BER of ; when thebroadcast service was enabled and the frequency offset betweenthe two downstream carriers was 2.5 GHz ( FSR of theDI), the BER performance of the unicast signal was degradedby the crosstalk from the broadcast signal and an error floorwas observed at a BER of . Although error-free trans-mission could be achieved by adopting forward error correction(FEC), it raises the cost of both OLTs and ONUs. To avoid thenecessity of employing FEC, the frequency offset was increasedto 7.5 GHz ( FSR of the DI), which helps to mitigate thecrosstalk-induced performance degradation. During this mea-surement, the relative bit delay between the two downstreamdata was adjusted to achieve the best BER performance. In thiscase, we could achieve a BER less than and the receiversensitivities were measured to be and dBm forthe downstream unicast and broadcast signals, respectively. Apower penalty of dB was incurred in the presence of theoffset polarization-multiplexed signals. Experiment and simu-lation results showed that the orthogonality between the unicastand broadcast signals is crucial for downstream detection. Theeye opening of the detected unicast (broadcast) signal was com-pletely closed due to the crosstalk from the broadcast (unicast)signal if the two parallel polarized downstream signals werecombined by an optical coupler instead of a PBC.It is interesting to note that further increasing the frequency

offset will reduce the power penalty; however, it will lower thespectral efficiency. The downstream unicast and broadcast car-

Fig. 5. BER performance of the downstream signals. The number inside theparenthesis in the legend indicates the frequency offset, .

riers for each ONU should be symmetric about the ITU gridfrequency and stay within one DWDM channel. This avoids al-locating additional wavelength channels for broadcast servicesin a DWDM system with a channel spacing of 50 GHz or even25 GHz. Fig. 6 shows the measured optical spectra of the offsetpolarization-multiplexed downstream signals with different fre-quency offsets. The resolution of the optical spectrum analyzerused in this measurement was 0.02 nm. The results clearly showthat the signal linewidth increased with the frequency offset.Considering the tradeoff between the downstream performanceand the spectral efficiency as well as the filtering effect of theAWG in the proposed architecture, we chose 7.5 GHz as thefrequency offset between the two downstream signals in the fol-lowing analyses. It is also worth noting that the receiver sensi-tivities of the downstream unicast (broadcast) signal can be im-proved if a balanced receiver is adopted since the broadcast (uni-cast) signal experiencing incoherent summation of two neigh-boring bits on both the constructive and destructive ports willbe mostly canceled out by the balanced receiver [8].For uplink, polarization dependence of the FP-LD requires

a depolarized seeding source if no polarization control is pro-vided at the ONU. The polarization-multiplexed downstreamsignals with a frequency offset of 7.5 GHz naturally meet this


Fig. 6. Measured optical spectra of (a) 10-Gb/s NRZ-DPSK signal and(b)–(d) the polarization-multiplexed 10-Gb/s NRZ-DPSK signals with fre-quency offset of 2.5, 7.5, and 12.5 GHz, respectively.

requirement and are reused as the seeding light to injection-lockthe FP-LD for upstream transmission. The DOP of the seedingsource was measured to be 2–4%. The depolarized seeding lightalso helps to reduce the RB-induced crosstalk in the single-fiber loopback configured network since the beating between thesignal and the RB is polarization sensitive [14]. Furthermore,the linewidth of the seeding light was widened due to the fre-quency offset between the two polarization beams, as shown inFig. 6. This also helps to reduce the RB-induced penalty [15].The signal-to-backscattering-power ratio in our experimentaldemonstration was measured to be around 29.5 dB. The up-stream BER performance is shown in Fig. 7. The receiver sensi-tivity at a BER of was dBm in the single-fiber loop-back configuration and dBm in the unidirectional con-figuration with two feeder fibers (refer to Fig. 4 for the setup).Therefore, the power penalty induced by the RBwas 1.2 dB. An-other advantage of the seeding source composed of NRZ-DPSKsignals is the relative constant intensity of phase modulation ex-cept for the transition-induced intensity dips from MZM-basedDPSK transmitters when compared to OOK modulation [16].There is no need to sacrifice the ER of the downstream sig-nals to constantly injection-lock the FP-LD [7]. The impact ofthe downstream modulation on the upstream performance isalso investigated through BER measurement. As shown in Fig.7, the receiver sensitivity of the upstream signal wasdBm when the downstream modulation was turned OFF but theseeding power was kept the same. Thus, the remodulation-in-duced power penalty was 1.8 dB. This should be ascribed tothe transition-induced intensity dip of the downstream signalsand can be reduced by using phase modulators for downstreamDPSK modulation.To demonstrate that the seeding light in our scheme could

constantly injection-lock the FP-LD without any polarizationcontrol, we deliberately added an electronically driven PC be-fore the FP-LD. The seeding light was randomly rotated every10 s before being fed into the FP-LD and the upstream BER was

Fig. 7. BER performance of the upstream signal.

Fig. 8. Measured upstream BERs versus time when the seeding light is ran-domly rotated every 10 s.

recorded. Fig. 8 shows the measured BERs at a received op-tical power of dBm. The results show a stable BER over10 min. The weak-resonant-cavity FP-LD, whose front-facetand rear-facet reflectivities are 1% and 99%, respectively, ex-hibits wider injection-locking range when compared to conven-tional FP-LDs. The BER performance of the upstream signalwas measured when the wavelength offset, , where

is the center wavelength of the offset polarization-multi-plexed downstream signals and is the wavelength of oneof the longitudinal modes of the free-running FP-LD nearby

, was tuned from 0 to 0.56 nm, which is the longitudinalmode spacing of the free-running FP-LD. The results are plottedin Fig. 9. The best receiver sensitivity of dBm at a BERof was obtained when the wavelength offset was 0.29 nmand this value is used as a reference in Fig. 9. The results showthat a BER of was achieved over the wavelength offsetranged from 0.10 to 0.41 nm despite that the maximum powerpenalty was 4.8 dB at the upper bound of this range.


Fig. 9. Power penalties at a BER of of the upstream signal versus wave-length offset.

Next, we analyze the power budget of the demonstratedWDM-PON system. For downlink, the seeding power for theFP-LD used in the experiment was kept at dBm and thereceiver sensitivity of the unicast or broadcast signal was about

dBm. Thus, the downstream power budget was primarilylimited by the required seeding power of the FP-LD. The totallink loss was 19 dB, including two AWGs (5.0 dB 2), 20-kmSSMF (4.0 dB), one 3-dB optical coupler, and other compo-nents (2.0 dB) such as PBC, circulator, and connectors. Thus,the launch power of the offset polarization-multiplexed down-stream signals should be at least 9.0 dBm, i.e., 6.0 dBm fromeach polarization beam, to meet the power requirement of bothdownstream detection and upstream remodulation. For uplink,the launch power of the injection-locked FP-LD was measuredto be 2.5 dBm. A power margin of 14.0 dB was obtained whenthe upstream receiver sensitivity was dBm.

IV. SIMULATION RESULTS

The robustness of the proposed scheme against the relativebit delay between the two downstream data, the frequency de-viation of the downstream signals from the DI, and the mis-matched interferometer delay is investigated through simulationin which the downstream DPSK signals were detected by a bal-anced receiver. In the following analyses, the balanced receiverwas tuned to detect one of the offset polarization-multiplexedsignals, the unicast signal. Thus, the broadcast signal, whichwas 7.5 GHz away in frequency, was regarded as crosstalk. Thesame analyses and conclusions made in this section also applyto the detection of the broadcast signal. The simulation was im-plemented by the commercial software OptiSystem 9.0 [17].Although the two offset polarization-multiplexed signals are

independent and orthogonal to each other in polarization, the de-tection of one signal would still experience the crosstalk fromthe other signal. This is because the crosstalk suppression of theDI could not be perfect throughout the entire bit duration, andthus, the relative bit delay between the unicast and broadcastdata affects the system performance. By tuning the relative bit

Fig. 10. Power penalty (at a BER of ) of the downstream unicast signalas a function of the relative bit delay between the unicast and broadcast data.

delay between the two signals, the peak of the crosstalk can belocated to different positions of the eye diagram of the detectedsignal. The deleterious effect of the crosstalk becomes largestwhen the peak of the crosstalk falls on the sampling instance ofthe detected signal. Fig. 10 shows the receiver power penaltyversus the relative bit delay between the two offset polariza-tion-multiplexed signals. It shows that the relative bit delay hada marginal effect on the downstream performance. The penaltieswere always less than 1.3 dB, which can eliminate the neces-sity of controlling the relative bit delay. Nevertheless, the worstperformance occurred when the two data were bit-aligned (i.e.,

or 100 ps). The best performance was obtained at therelative bit delays of 30 and 70 ps. At these bit delays, we hadslightly better performance than when the two data were half-bitdelayed. We ascribe this to the small asymmetry of signal eyes,which is caused by the electrical filter at the receiver. The cor-responding eye diagrams are included in the inset in Fig. 10.In DPSK reception, the deviation of the laser frequency from

the DI leads to performance degradation [9]. Fig. 11 depicts thefrequency deviation-induced power penalties. The dash curverepresents the case when the unicast signal was exactly alignedto the penalty-free frequency (i.e., ) but the broadcastsignal deviated from the infinite-penalty frequency [i.e.,

]; the solid curve represents the case when thebroadcast signal was exactly aligned to the infinite-penalty fre-quency but the unicast signal deviated from the penalty-free fre-quency. It shows that the performance of the unicast signal wasmuch more sensitive to the deviation of the broadcast signal.For a penalty of less than 1.0 dB, the frequency deviation of thebroadcast signal should be kept below 250MHz. This frequencyoffset tolerance could be achieved by using temperature-stabi-lized DIs. A feedback servo loop would be required to stabilizethe DI and track the laser wavelength drift.Another aspect closely related to the frequency deviation

is the polarization dependence of the DI, which can be char-acterized by polarization-dependent wavelength shift .

is the wavelength shift of the DI wavelength response de-pending upon the state of polarization (SOP) of the input signal.


Fig. 11. Power penalty (at a BER of ) of the downstream unicast signalversus the frequency deviation of the downstream signals from the DI.

Fig. 12. Power penalty (at a BER of ) of the downstream unicast signalversus the interferometer delay-to-bit-rate mismatch.

Although the offset polarization-multiplexed downstream sig-nals may remain orthogonally polarized during transmission,the SOP of each signal is generally unknown and changes overtime. Since no polarization control is provided at the receiverside in our scheme, the power penalty induced by the polariza-tion dependence of the DI is inevitable. As discussed before,the deviation of the broadcast signal has a stronger effect onthe detection of the unicast signal, so the worst performance ofthe unicast signal detection caused by polarization dependenceof the DI occurs when the broadcast signal deviates from theinfinite-penalty frequency for an amount of . For a penaltyof less than 1.0 dB, the polarization-dependent frequencyshift of the DI should be kept below 250 MHz, and therefore,the should be smaller than 2 pm. Most commerciallyavailable 10-GHz DIs satisfy this requirement [18], [19].Fig. 12 shows the receiver sensitivity degradation induced

by the mismatch between the interferometer delay and the datarate of the downstream signals. The mismatch is denoted by theratio between the interferometer delay and the data rate. When

the broadcast service was disabled, a 5% mismatch led to lessthan 0.1-dB power penalty. However, when the broadcast ser-vice was enabled, the power penalty induced by the interferom-eter delay-to-bit-rate mismatch increased dramatically. In fact,the interferometer delay-to-bit-rate mismatch is closely relatedto the frequency deviation between the broadcast signal andthe infinite-penalty frequency. A 5% mismatch could lead to a375-MHz deviation of the broadcast signal at maximum, whichgave rise to around 4.0-dB power penalty as shown in Fig. 11.To control the power penalty within 1.0 dB, the interferometerdelay-to-bit-rate mismatch should stay within 3%. Imperfectionin DI manufacturing may cause such a mismatch; however, theerror of the interferometer delay of commercial DIs is usuallysmaller than 1% [18].

V. CONCLUSION

We have proposed a WDM-PON system with a downstreambroadcast overlay using the offset polarization multiplexingtechnique. The downstream DPSK-formatted unicast andbroadcast signals, which are separated from each other bya quarter or three quarters of the data rate in frequency, arepolarization-multiplexed at the transmitter. The offset polariza-tion-multiplexed signals are demultiplexed and demodulatedby a pair of DIs without any polarization tracking at the re-ceiver. The offset polarization-multiplexed downstream signalsare also exploited to facilitate the constant injection-lockingof polarization-sensitive FP-LDs for upstream transmission.Therefore, no polarization control or tracking is needed forboth downstream and upstream transmissions. Based on theproposed scheme, we have successfully demonstrated thebidirectional transmissions of 10-Gb/s unicast and 10-Gb/sbroadcast signals for downlink and 2.5-Gb/s signal for uplink.Our experiment and simulation studies show that the proposedscheme is very robust against the polarization fluctuation on thelink and the relative bit delay between the downstream unicastand broadcast signals. The simulation study also reveals that thedownstream performance is sensitive to the frequency deviationof the downstream signals from the DI. However, the frequencydeviation can be controlled within an acceptable range withcommercially available DIs and wavelength-stabilized lasersources. A feedback servo loop would be needed to stabilizethe DIs and track the laser wavelength drift at the ONU.

REFERENCES[1] J.-H. Moon, K.-M. Choi, and C.-H. Lee, “Overlay of broadcasting

signal in a WDM-PON,” presented at the Optical Fiber Communica-tion Conf., Anaheim, CA, 2006, paper OThK8.

[2] F. Zhang, W.-D. Zhong, Z. Xu, T. H. Cheng, C. Michie, and I. An-donovic, “A broadcast/multicast-capable carrier-reuseWDM-PON,” J.Lightwave Technol., vol. 29, no. 15, pp. 2276–2284, Aug. 2011.

[3] Y. Tian, Q. Chang, and Y. Su, “A WDM passive optical network en-abling multicasting with color-free ONUs,” Opt. Express, vol. 16, pp.10434–10439, Jun. 2008.

[4] Y. Zhang, N. Deng, C.-K. Chan, and L.-K. Chen, “A multicastWDM-PON architecture using DPSK/NRZ orthogonal modulation,”IEEE Photon. Technol. Lett., vol. 20, no. 17, pp. 1479–1481, Sep.2008.

[5] J. Xu, Z. Liu, L.-K. Chen, and C.-K. Chan, “Time-interleaved phaseremodulation to enable broadcast transmission in bidirectional WDM-PONs without additional light sources,” presented at the Optical FiberCommunication Conf. and Exposition, Los Angeles, CA, 2011, paperOThK4.


[6] F. Xiong, W.-D. Zhong, and H. Kim, “A WDM-PON enabling broad-cast service based on polarization multiplexing,” presented at the Conf.Lasers and Electro-Optics/Pacific Rim, Sydney, 2011, 3440-CT-4.

[7] F. Xiong, W.-D. Zhong, and H. Kim, “A broadcast-capableWDM-PON based on polarization-sensitive weak-resonant-cavityFabry-Perot laser diodes,” J. Lightwave Technol., vol. 30, no. 3, pp.355–361, Feb. 2012.

[8] H. Zhang, J.-X. Cai, C. R. Davidson, B. Anderson, O. Sinkin, M.Nissov, and A. N. Pilipetskii, “Offset PDM RZ-DPSK for 40 Gb/slong-haul transmission,” presented at the Optics Fiber Communica-tion, San Diego, CA, 2009, paper OThR2.

[9] H. Kim and P. J. Winzer, “Robustness to laser frequency offset in di-rect-detectionDPSK andDQPSK systems,” J. Lightwave Technol., vol.21, no. 9, pp. 1887–1891, Sep. 2003.

[10] K. Higuma, S. Oikawa, Y. Hashimoto, H. Nagata, and M. Izutsu,“X-cut lithium niobate optical single-sideband modulator,” Electron.Lett., vol. 37, pp. 515–516, Apr. 2001.

[11] K. Voigt, L. Zimmermann, G. Winzer, M. Schnarrenberger, T. Mitze,J. Bruns, and K. Petermann, “Silicon-on-insulator (SOI) delay-line in-terferometer with low polarization-dependent wavelength shift,” pre-sented at the Eur. Conf. and Exhibition of Optical Communication,Copenhagen, Denmark, 2007, paper ThC2.

[12] G. Unterborsch et al., “Hybrid flip-chip integration of a 40 Gb/s DPSKreceiver comprising a balanced photodetector on a DLI-SOI board,”presented at the Eur. Conf. Optical Communication, Brussels, Bel-gium, 2008, paper P.2.04.

[13] P. J. Winzer and J. Leuthold, “Return-to-zero modulator using a singleNRZ drive signal and an optical delay interferometer,” IEEE Photon.Technol. Lett., vol. 13, no. 12, pp. 1298–1300, Dec. 2001.

[14] M. O. van Deventer, “Polarization properties of rayleigh backscat-tering in single-mode fibers,” J. Lightwave Technol., vol. 11, no. 12,pp. 1895–1899, Dec. 1993.

[15] J.-H. Moon, K.-M. Choi, S.-G. Mun, and C.-H. Lee, “Effects of back-reflection inWDM-PONs based on seed light injection,” IEEE Photon.Technol. Lett., vol. 19, no. 24, pp. 2045–2047, Dec. 2007.

[16] A. H. Gnauck and P. J. Winzer, “Optical phase-shift-keyed transmis-sion,” J. Lightwave Technol., vol. 23, no. 1, pp. 115–130, Jan. 2005.

[17] OptiSystem 9.0, [Online]. Available: http://www.optiwave.com/prod-ucts/system_features.html

[18] Optoplex DPSK Demodulator, [Online]. Available: http://www.opto-plex.com/DPSK_Demodulator.htm

[19] Avensys Tech., [Online]. Available: http://www.avensystech.com/AvensysTech/Products.php?locale=en&Line_no=13&sub_cate-gory_id=155&cat_category_id=138

Fei Xiong (S’09) received the B.S. degree in electrical engineering from theBeijing University of Posts and Telecommunications, Beijing, China, in 2009.She is currently working toward the Ph.D. degree in School of Electrical andElectronic Engineering, Nanyang Technological University, Singapore.Her research interest includes wavelength-division-multiplexed passive op-

tical networks.

Wen-De Zhong (SM’03) received the Ph.D. degree from the University ofElectro-Communications, Tokyo, Japan, in 1993.He is a Professor in the School of Electrical and Electronic Engineering,

Nanyang Technology University (NTU), Singapore. He has made significantcontribution in the area of optical communication systems and networks, andpublished more than 200 journal and conference papers. He has given severalinvited presentations at international conferences. He has served on organizingand/or technical program committee or advisory committee for numerous in-ternational conferences. During 1993–1995, he was a Postdoctoral Fellow atNTT Network Service Systems Laboratories, Tokyo. From 1995 to 2000, hewas with the Department of Electrical and Electronic Engineering, University ofMelbourne, Australia, as a Research Fellow and then a Senior Research Fellowduring 1998–2000. He joined NTU in 2000 where he is leading a team involvedin research on optical communication systems and network. His research in-terests include optical WDM systems and networks, optical signal processing,photonic switching systems, and network survivability.

Hoon Kim (SM’11) received the M.S. and Ph.D. degrees in electrical en-gineering from the Korea Advanced Institute of Science and Technology(KAIST), Taejon, Korea, in 1996 and 2000, respectively.From 2001 to 2002, he was with Bell Laboratories, Lucent Technologies,

where he was involved in research on advanced optical modulation formats.From 2002 to 2007, he was with Samsung Electronics, Suwon, Korea, wherehe was involved in research on optical duobinary transmission systems, opticalsources for access applications, and fiber-optic networks for wireless commu-nications. Since 2007, he has been an Assistant Professor in the Departmentof Electrical and Computer Engineering, National University of Singapore,Singapore.Dr. Kim now serves as an Associate Editor of the IEEE PHOTONICS

TECHNOLOGY LETTERS and has served on several international conferences asa member of technical committee, including Optical Fiber Communication,IEEE Lasers and Electro-Optics Society, OptoElectronics and CommunicationConference, and Asia-Pacific Optical Communications.

a broadcast-capable wdm passive optical network using offset polarization multiplexing

Documents