Multiplexing and Direct Detection
Dayou Qian,Neda Cvijetic ,Member,IEEE ,Junqiang Hu,and Ting Wang
Abstract—In this paper,we propose and experimentally demon-strate the first
single-40Gb/s and 108Gb/s multiple-input multiple-output orthogonal frequency-division multiple access (OFDMA)passive optical networks (PON)architecture for next-generation PON systems based on OFDM,polarization multiplexing (POLMUX),and direct detection.Superior perfor-mance was exhibited after 20km SSMF transmission and a 1:32optical split.The novel POLMUX approach greatly simplified receiver-end hardware compared to coherent detectors,while increasing spectral efficiency to enable 40+Gb/s data rates.Moreover,the proposed solution achieved the highest single-wave-length downstream transmission reported to date in any PON system.As such,the introduced architecture may be viewed as a highly attractive candidate for next-generation optical access.Index Terms—Direct detection,orthogonal frequency-division multiple access (OFDMA),passive optical networks (PON),polar-ization multiplexing.
I.I NTRODUCTION
F
UELED by an exponentially growing demand for broad-band services and the advent
of
Gb/s Ethernet tech-nologies,the R&D focus for passive optical networks (PON)has shifted toward next-generation access systems that are capable of providing a “future-proof”broadband solution [1]–[3].How-ever,it is also critical that these optical access systems be highly flexible and cost-efficient to readily accommodate emerging ser-vices and applications [2].Consequently,unlike long-haul net-works where short-term flexibility is not a primary goal and dis-tance-bandwidth products are sufficiently large to leverage high
implementation cost,access networks
(
km)must be easily reconfigurable and maintain low hardware and operational com-plexity to remain attractive and practical.
It is well known that advanced modulation formats are a cost-effective way to increase the transmission data rate.Due to its resilience to both chromatic and polarization-mode disper-sion,spectral efficiency,and natural compatibility with digital signal processing (DSP)-based implementation,optical orthog-onal frequency-division multiplexing (OFDM)has emerged as an attractive candidate for future optical transmission systems [4]–[8].Moreover,in the context of PON-based optical access,our recently proposed orthogonal frequency-division multiple access (OFDMA)PON solutions can be used to transparently support various applications and enable dynamic bandwidth
Manuscript received May 01,2009;revised July 28,2009.First published August 07,2009;current version published February 01,2010.
The authors are with NEC Laboratories America,Princeton,NJ 08540USA (e-mail:dqian@nec-labs.com;neda@nec-labs.com;jqhu@nec-labs.com;ying@nec-labs.com).
Digital Object Identifier 10.1109/JLT.2009.2029541
allocation among them [9],[10].In this way,OFDM-PON and OFDMA-PON [9]–[12]provide a novel transmission and networking paradigm that enables both cost-effective spectral efficiency and dynamic bandwidth granularity for next-gener-ation optical access networks.
The performance of both directly detected [6]–[8],[13]–[16]and coherent optical (CO)OFDM [5],[17]–[21]has received great attention recently.In coherently detected optical OFDM transmission,polarization multiplexing (POLMUX),wherein a high-speed OFDM signal is carried in each of two orthogonal polarizations,has been widely proposed and demonstrated as an excellent way to further increase spectral efficiency at ultra-high speeds [5],[21].The tradeoff for higher spectral efficiency in such multiple-input multiple-output (MIMO)POLMUX sys-tems is the increased complexity mandated by coherent detec-tion,which entails narrow linewidth lasers for receiver-end local oscillators (LOs),as well as complex frequency-offset and phase noise compensation DSP algorithms.Due to data rate limita-tions imposed by current electronic digital-to-analog converters (DAC),increasing spectral efficiency via POLMUX would also be of great value in OFDM-based networks as well [22],[23].However,the inherent complexity of coherent receiver-end ar-chitectures restricts their use in cost-sensitive access and metro application.
In this work,we provide an extended description of our recently proposed and experimentally demonstrated directly detected POLMUX-OFDMA-PON [23],[24],whereby two polarization-orthogonal OFDMA signals are recovered by direct detection (DD)on two photodiodes,followed by post-detection DSP.To the best of our knowledge,these results document the first DD-based method for POLMUX optical OFDM transmission.The remainder of the paper is organized as follows.In Section II,the principles and key benefits of OFDM-PON optical access systems are introduced.Section III details the challenges of POLMUX with DD (POLMUX-DD),while Section IV introduces the architecture of our proposed solution,which circumvents these difficulties.Novel channel estimation algorithms for POLMUX-DD OFDMA-PON are presented in Section V.Section VI contains the experimental setup and results for both 40Gb/s and 108Gb/s POLMUX-DD OFDMA PON transmission,while Section VII summarizes and concludes the paper.
II.OFDMA-PON P RINCIPLES
The continuing increase in bandwidth demand in access networks has spurred the emergence of several next-gen-eration PON systems,including time-division multiplexing (TDM)-based 10G-PON and 10GE-PON and wavelength
0733-8724/$26.00©2010IEEE
Fig.1.OFDMA PON architecture for delivery of heterogeneous services.
division multiplexed (WDM)-PON [9].However,migration to these new technologies also raises several challenges.For example,to protect legacy network investments,there is in-terest in providing next-generation services primarily through TDM-based techniques;yet,TDM-based 10G-PON and 10GE-PON mandate expensive 10Gb/s components,complex scheduling algorithms and framing technology,and are highly sensitive to packet latency [9].Moreover,looking beyond 10Gb/s PON transmission,the practical design and implementa-tion
of
Gb/s burst-mode TDM receivers currently remain a daunting task from both the technical and economic perspec-tives [9].WDM-based approaches also feature a prohibitive cost barrier due to the need to fundamentally alter legacy PON distribution networks,and adopt either colored or wave-length-tunable ONU-side equipment,respectively.Moreover,WDM-PON technology is also unable to dynamically allocate bandwidth with
sub-granularity,which significantly reduces flexibility.Likewise,WDM-PON still lacks the flexibility to dynamically allocate the bandwidth among multiple services and raises system cost due to requirements for multiple trans-ceivers.Consequently,a cost-effective,flexible scheme that addresses these challenges is of great interest.
We have recently proposed a novel PON architecture by using OFDMA,shown in Fig.1,to transparently support various applications and enable dynamic bandwidth allocation among them [9],[10].The OFDMA-PON approach illustrated in Fig.2is essentially a hybrid technique,which combines OFDM and TDMA,such that the OFDM subcarriers can be dynamically assigned to different services in different time slots,as illustrated in Fig.2.While this approach can increase complexity of uplink multiple access protocols,it also offers very wide bandwidth flexibility.Moreover,as demonstrated in [11]and [12],OFDM-based PON solutions readily lend them-selves to adaptive per-subcarrier modulation and constellation expansion that may be used to cost-effectively upgrade the data
rate of existing 1Gb/s networks
to
Gb/s.As shown in Fig.1,in the proposed OFDMA-PON system architecture,dedicated subchannels,which are composed of one or more OFDM subcarriers,become transparent pipes for delivery of arbitrary analog or digital signals for both circuit-and packet-switched systems.For example,dedicated subcarriers (white and black OFDM subbands in Fig.1)
can
Fig.2.Frequency-and time-domain partitioning of an OFDMA frame.Dif-ferent colors of time/frequency blocks denote resources assigned to different services.Each ONU will select those time/frequency slots that have been pre-assigned to it according to the schedule distributed from the OLT.
be reserved as orthogonal,transparent pipes for legacy TDM (T1/E1)services and RF mobile base station signals,respec-tively.For downstream traffic,the OLT can reserve some subcarriers as dedicated,transparent pipes and encapsulate packet-based data into remaining OFDM bands and time slots,according to the specific frequency-and time-domain scheduling results.The OFDM frame and other analog signals are next mixed by an electrical coupler to drive the optical modulator.At the ONU side,each ONU selects its own data or signal from its preassigned subcarrier(s),pipes and time slots,as communicated by the OLT scheduler.To transmit upstream traffic,each ONU maps its data and/or signal to its assigned OFDM subcarrier(s),nulls all remaining subcarriers,and per-forms OFDM modulation to generate a complete frame.With this approach,the OFDMA-PON is both flexible and extensible to any emerging/future applications.While the upstream archi-tecture is also of great interest,in this paper,we will focus on high-speed downstream transmission in OFDMA-PON using our our newly developed technique of polarization multiplexing with direct detection (POLMUX-DD).
III.P OLARIZATION M ULTIPLEXING W ITH D IRECT D ETECTION In both coherently and directly detected optical OFDM sys-tems,the optical OFDM signal is generated through subcar-rier multiplexing of a multi-Gigahertz electrical OFDM signal onto an optical carrier.Consequently,since the electrical OFDM signal can only be generated by high-speed DAC,current DAC
In order to implement POLMUX-DD,a challenge that is not present in coherent-detected POLMUX transmission mustfirst be addressed.Namely,in coherent systems[5],[17]–[21],the orthogonal polarization components of the LO signal are lo-cally known and can be perfectly separated using a polarization beam splitter(PBS).In this way,after coherent mixing with the incoming optical signal,full postphotodetection separation of the polarization-multiplexed data bands can readily be achieved.Under DD,however,the optical carrier that travels with the optical signal is also used as the LO;consequently, the LO polarization state at the PBS input is no longer known. As a result,the LO polarization components at the PBS output will each contain a mixture of the two original LO polarization states as defined at the transmitter,such that the LO will not be perfectly orthogonal to either of the POLMUX data bands. The subsequent beating of nonorthogonal LO components with the POLMUX signal will cause destructive interference among the POLMUX data bands.At the photodetector output,this interference will translate to partial or complete signal fading. An example of this cross-polarization interference effect is shown in Fig.3(a)(d).In Fig.3(a),a single-sideband(SSB) optical OFDM signal is shown with its optical carrier prior to POLMUX.In Fig.3(b),the optical SSB-POLMUX-OFDM signal shown on the left passes through an arbitrarily oriented PBS,resulting in a mix of input polarization components at each PBS output.Due to DD,the mixed polarization compo-nents combine incoherently,and since polarization information is lost during this process,the interference effect cannot be undone and the signal fading remains.Moreover,in another potential scenario shown in Fig.3(c),the local polarization state of the LO signal might be exclusively matched to one of the PBS branches and cause the complete loss of one of the POLMUX data bands.Most recently,an architecture that exploits self-coherent detection to obviate this problem [Fig.3(d)]was proposed and demonstrated in[25],[26],yet this solution also mandates a coherent receiver architecture that is not cost-feasible for access and metro systems.In the following section,we will introduce our POLMUX-DD archi-tecture that solves the cross-polarization interference problem described here,while maintaining the cost-efficient structure of a traditional DD receiver.
IV.P ROPOSED P OLMUX-DD S YSTEM A RCHITECTURE
As noted in Section III,the inability to preserve polariza-tion orthogonality between the LO and OFDM sidebands at the POLMUX receiver causes cross-polarization
beating Fig.3.(a)SSB optical OFDM signal with optical carrier;(b)SSB-POLMUX-OFDM signal before/after DD;(c)SSB-POLMUX-OFDM signal with sideband fading;(d)Self-coherent reception of a SSB-POLMUX-OFDM signal.
and destructive interference in straightforward extensions of the coherent POLMUX approach to DD systems.In the novel POLMUX-DD approach proposed here,however, such nonorthogonal LO/signal beating is sidestepped by exploiting two frequency-orthogonal optical carriers at the POLMUX-OFDM transmitter,one for each POLMUX band. Since frequency orthogonality is preserved during the transmis-sion process,DD can be used to fully recover the transmitted signal in each polarization,without the need to know the local polarization state of the LO.Following such interfer-ence-free photodetection,novel MIMO channel estimation and equalization techniques need to be used to recover the polarization-specific baseband data symbols.(The discussion of the new MIMO algorithms is reserved for Section V.) Fig.4shows the schematic diagram of the proposed MIMO-OFDM-PON featuring POLMUX and DD.At the OLT,a con-tinuous-wave(CW)laser drives an intensity modulator(IM), which is modulated by a clock source with carrier suppression, generating two optical carriers X and Y separated by two times the clock source
frequency,.As will be shown in the experi-mental setup of Figs.10and11,for example,
when
GHz,the two optical carriers are25GHz apart in the frequency domain.By carefully choosing the wavelength of the CW laser, a50GHz optical interleaver can be used to separate the two op-tical carriers.It is noted that in the frequency domain descrip-tion of Fig.4,the optical carrier denoted by the dashed lines isFig.4.Proposed POLMUX-DD architecture(top)and frequency-domain description of the SSB-POLMUX-OFDM signal generation(bottom).
in factfiltered out by the optical50GHz interleaver and is only shown as a reference to explain the frequency domain place-ment of the modulated OFDM signals.Each individual optical carrier next drives a separate IM,modulated by an independent RF OFDM signal,which is generated by upconversion of the baseband OFDM signal via the IQ-mixer.The RF carrier fre-quency of the IQ mixer is chosen to be the same as that of the clock source modulating thefirst IM.
Next,the two IM outputs are combined with a polarization beam combiner(PBC)to generate a POLMUX-OFDM signal with dual POLMUX carriers having orthogonal polarizations, as shown in Fig.4(bottom,fourth insert).Finally,the two outer OFDM side bands arefiltered out with a25GHz optical in-terleaver and the SSB-POLMUX-OFDM output signal is sent downstream.The example PON architecture of Fig.4adopts a20km transmission range,with an additional1:32splitter to emulate ONU-s within the20km distance.
At the receiver side,the POLMUX-OFDM signal is divided by a PBS,and the two PBS outputs are directly detected by two separate photodiodes.At this stage,the OFDM signals are still RF signals,so following analog-to-digital conversion (ADC),they are downconverted to the baseband by two OFDM receivers,which output frequency-domain data samples that have contributions from both original polarizations.Finally,the MIMO polarization demultiplexing(PolDeMux)receiver re-covers the original data in each polarization via DSP algorithms described next.We also note that since the proposed PolDeMux algorithms are designed for an arbitrary incoming polarization state,variations among ONU-side polarization states will not affect algorithm performance.
V.N OVEL C HANNEL E STIMATION A LGORITHM FOR
P OLMUX-DD S YSTEMS
In POLMUX system with coherent detection,complete po-larization information is preserved during photodetection,such that from the channel estimation and equalization point of view, the problem is entirely analogous to a two-input,two-output MIMO system pioneered and well-documented in wireless liter-ature[27]and subsequently extended to thefiber-optic channel in[5],[21].Moreover,another reason why the
22MIMO extension can readily be made for coherently detected optical OFDM is that both polarizations are detected with respect to the same optical carrier.The equalization task thus consists of derotating polarization components based on the preserved po-larization information and with respect to the same optical car-rier reference.
In the POLMUX-DD system proposed here,however,polar-ization information is not preserved and two independent op-tical carriers are needed to prevent cross-polarization-induced signal fading at the OFDM demodulator input.From the channel equalization perspective,the need for two frequency-orthog-onal carriers introduces two different optical carrier references. The received POLMUX signal thus experiences two different
22MIMO optical channels:one with respect to each indepen-dent optical carrier.Moreover,since the POLMUX-OFDM data bands overlap in the frequency domain(Fig.4),the different
22MIMO channels cannot be considered separately in the channel estimation and equalization process but must be treated in a joint fashion as a single
44MIMO channel.The channel estimation problem for the proposed POLMUX-DD system thus consists of obtaining coefficients of this
44matrix.Since this is a situation not previously encountered in the context of CO OFDM,new training signal patterns,channel estimation algo-rithms,and PolDeMux approaches are needed.
Before introducing the new training sequences used for channel estimation,wefirst set forth the channel model and notation used throughout this section in Fig.5.As shown by Fig.5,due to polarization rotation,signal components on each of the two input polarization states,Pol-X and Pol-Y,may migrate to one of the output polarization states,Pol-X′and Pol-Y′.The notation XY’,for example,denotes component migration from input polarization Pol-X to output polarization Pol-Y′.Moreover,in Fig.5,
coefficients
and are used to denote the polarization rotation of the OFDM data band and optical carrier signals,respectively.Thus,denotes
Fig.5.Channel model for POLMUX-OFDM transmission with two fre-quency-orthogonal optical
carriers.
Fig. 6.Detailed frequency domain description of SSB-POLMUX-OFDM signal.
the channel coefficient experienced by the OFDM data band component that migrates from Pol-X to Pol-Y ′,
while
refers to the coefficients decided by the portion of the optical carrier launched from Pol-X that ends up in Pol-Y ′.We note that frequency-flat fading is assumed over the range of interest,which is valid for PON bandwidths and transmission distances of interest [23],[24].
A detailed frequency domain description of the optical SSB-POLMUX-OFDM signal and the associated notation used in channel estimation and equalization tasks are shown in Fig.6.(We note that a qualitative description of Fig.6is also
given in the bottom rightmost insert of Fig.4.)In Fig.
6,
and denote frequency-orthogonal optical carriers launched in Pol-X and Pol-Y ,
respectively,is the source clock frequency
(Fig.4),
and
denotes the frequency-domain separation between the optical carriers and the innermost edge of the
POLMUX-OFDM bands.The
notation
denotes the com-plex baseband OFDM symbol modulated in Pol-X onto the th
OFDM
subcarrier;
carries an analogous meaning,
where is the fast Fourier transform (FFT)size.In a similar
manner,
and ,respectively,denote the complex OFDM symbols on the th
and
th subcarrier in Pol-Y.
The novel training pattern,shown in Figs.7and 8,was com-posed of two different training sets:A and B.Training set A,
shown in Fig.7,consists of
subsets
and during which known symbols are transmitted on OFDM subcarriers in the fre-quency
range
in Pol-X and Pol-Y ,respectively,where the
range
is defined with respect to the polarization-par-allel optical carrier.The training
subsets
and were trans-mitted sequentially in time,and their relative frequency-domain placement is also shown in Fig.6.During training set A,all OFDM subcarriers outside of the frequency range shown
in
Fig.7.Training set A:in Pol-X (left)and in Pol-Y (right).
Fig.7were nulled.In Fig.7,for example,transmitted
symbols
and belong in the
appropriate range in Pol-X and Pol-Y ,respectively.
Symbols
and ,de-noted by dashed rectangles in Fig.7,are outside of
the
frequency range with respect to the carrier in their launched po-larizations and are thus set to zero.However,due to polariza-tion rotation of the optical signal,components of the transmitted
symbols
and will appear in these dashed locations,ac-cording to the channel model of Fig.5,and will thus need to be accounted for as part of channel estimation.
Fig.8illustrates training set B,composed of
subsets
and ,and formed by transmitting training symbols on OFDM
frequencies
where
as referred to the optical carrier in Pol-X and Pol-Y ,respectively.All remaining OFDM
subcarriers are nulled in training set B,and
subsets
and are transmitted sequentially in time.The frequency-domain
placement
of
and is also illustrated in Fig.6.Again,due to polarization rotation governed by the channel model of
Fig.5,components of the transmitted
symbols
and will appear on zeroed
symbols
and ,re-spectively,as denoted by the dashed lines in Fig.8.The com-plete training signal for the proposed POLMUX-DD architec-ture consisted of at least one pair of training sets A and B trans-mitted sequentially in time;equivalently,the complete training
signal consisted of the time
sequence:
.Additional details about the insertion of the training sequence into the data stream are also given in the top right inserts of Figs.10and 11,for the 40Gb/s and 108Gb/s experiments.At the receiver end,symbol measurements were made in each of the output polarizations Pol-X ′and Pol-Y ′.We denote the
measured symbols on Pol-X ′
by
and ,corre-sponding to data measured on the th
and
th OFDM subcarriers.The corresponding notation for measured symbols
on Pol-Y ′
is
and .Gathering the data provided by training sets A and B and the received symbol measurements,we may form the
44channel estimation matrix
as
(1)
The PolDeMux matrix used in the MIMO-OFDM DSP re-ceiver(Fig.4)to undo polarization rotation and recover base-band data on each OFDM subcarrier may then be readily ob-tained
as
(2)
To illustrate the practical functionality of the new channel
estimation technique,an example is shown in Fig.9(a)and
(b),based on the experimental parameters for the108Gb/s
POLMUX-OFDMA-PON system.In Fig.9(a),the optical
spectrum of the108Gb/s SSB-POLMUX-OFDM signal is
shown to illustrate that,in this case,for OFDM subcarriers
labeled1–4in each polarization Pol-X and Pol-Y in Fig.9(a),
which are not frequency overlapped with data in the orthogonal
polarization,traditional MIMO PolDeMux algorithms based
on a
22channel matrix can be used.The reason for this is
that subcarriers1–4in Pol-X are frequency-domain-orthog-
onal to their counterparts in Pol-Y and vice versa.However,
for OFDM subcarriers5–8in Fig.9(a),which are subject to
cross-polarization overlap,the novel training estimation algo-
rithm proposed earlier must be used.The need for the novel
channel estimation approach is further highlighted in Fig.9(b),
which shows the electrical signal following DD of the optical
spectrum in Fig.9(a).We also note that the approach taken to
achieve40Gb/s MIMO-OFDM-PON transmission is directly
analogous to the example provided here,with the exception
that only subcarriers5–8were used in each polarization.
VI.E XPERIMENTAL S ETUP AND R ESULTS
Figs.10and11depict the two respective experimental setups
used to achieve40Gb/s and108Gb/s MIMO-OFDM-PON
downstream transmission over20km SSMF and an additional
15dB attenuation.The additional attenuation was inserted to
emulate a1:32optical split that is part of the class optical
distribution
network.
Fig.9.(a)Optical spectrum of the108Gb/s SSB-POLMUX-OFDM signal;
(b)Received electrical signal after DD of spectrum in Fig.9(a).
A.40Gb/s MIMO-OFDM-PON Transmission
Fig.10depicts the experimental setup for40Gb/s POLMUX-
OFDMA-PON downstream transmission with DD[23].At the
OLT transmitter,an OFDM baseband signal was generated off-
line,with16QAM used to map bit stream data onto each OFDM
subcarrier.The FFT
size and a1/32cyclic prefix(CP)
were applied.A training sequence was added every128OFDM
data frames.Thefirst25and last24OFDM subcarriers were
set to zero and the signal was upsampled by a factor of1.39.
The baseband OFDM signal was uploaded into a5GHz Tek-
tronix AWG7102arbitrary waveform generator(AWG)oper-
ating at10Gsample/s and8-bits DAC resolution,producing a
2.8GHz I/Q output.Next,an analog IQ-mixer was used to up-
convert the baseband OFDM signal to a12.5GHz RF.An ex-
ternal cavity laser(ECL)with wavelength of1549.844nm and
100kHz linewidth was employed as the CW optical source,and
was modulated with the same12.5GHz clock signal used in the
IQ-mixer.Subcarrier suppression was exploited to generate two
optical carriers with25GHz separation[Fig.10(a)].A50GHz
optical interleaver was employed to separate the two optical car-
riers,with each optical carrier used to drive an IM.Each IM was
modulated by the RF OFDM signal with a bandwidth of5.6
GHz centered at12.5GHz.The optical spectrum after the mod-
ulator is shown in Fig.10(b)and(c).The two resulting optical
OFDM signals were combined by a PBC,placing them on two
orthogonal polarizations and generating the optical spectrum of
Fig.10(d).A25GHz optical interleaver with35dB channel
Fig.10.Experimental setup and representative spectra of 40Gb/s MIMO-OFDMA-PON transmission using POLMUX-DD.
isolation was used to generate the final SSB-POLMUX-OFDM signal shown in Fig.10(e).
The SSB-POLMUX-OFDM signal was next amplified and transmitted through 20km fiber and one 15dB attenuator (equal to a 1:32splitter)to the ONU.The fiber used in the experiment was standard SMF-28fiber with 17ps/nm/km dispersion and an insertion loss of 0.2dB/km at 1550nm.The total downstream transmitted power was 8dBm.
At the ONU,the received power varied
between
dBm
and dBm,accounting for a 4–5dB attenuation from 20km fiber and the 15dB attenuator loss.In order to satisfy oper-ational requirements of the high-speed (20GHz)photodiodes,an erbium-doped fiber amplifier (EDFA)was placed before the PBS to adjust the optical power.The SSB-POLMUX-OFDM signal was next separated by the PBS and directly photode-tected by two 20GHz linear photodiodes.The received RF OFDM signals were sampled by a Tektronix real-time oscil-loscope (DPO72004)at 50Gsample/s,with all subsequent DSP done off-line.The OFDM receiver consisted of a digital IQ-demux,followed by an FFT.Both OFDM receiver outputs were fed to a MIMO PolDeMux receiver,which performed both channel estimation (enabled by training signals)as well as PolDeMux (using the channel estimation results),as described in Section IV.
The total overhead in the experiment was divided as follows:7%was used for forward error correction (FEC)coding,3.125%for CP insertion,and 0.78%for preambles.It is noted that,while the total transmission rate before coding was 44.8Gb/s,the post-coding data rate is 40Gb/s.
B.108Gb/s MIMO-OFDM-PON Transmission
In Fig.11,the experimental setup for 108Gb/s directly detected POLMUX-OFDMA-PON downstream transmission is illustrated [24].At the OLT transmitter,two independent 14.4Gb/s OFDM baseband signals and two independent 40Gb/s OFDM baseband signals were generated off-line and output continuously by two Tektronix AWGs,AWG7102(AWG1)and AWG7122B (AWG2)with sampling rates of 10Gsymbols/s and 12Gsymbols/s,respectively.For the 14.4
Gb/s OFDM baseband signals,the FFT size
is
,with 252data-bearing subcarriers,while for the 40-Gb/s OFDM
baseband
signals,
,with 200data-bearing subcarriers.In both cases,16QAM was used for symbol mapping.The 14.4Gb/s signals were digitally upconverted to a 1.85GHz intermediate frequency (IF),upsampled to 10Gsample/s,and output by AWG1directly.The 40Gb/s OFDM signals were upconverted to 10GHz IF in an analog fashion,with an elec-trical coupler used to combine the two IF bands and generate the electrical spectrum shown in Fig.11(a).After the digital upconversion,the IF OFDM signal frame length was doubled,with a corresponding 1/32CP per OFDM frame.Moreover,an extra 0.8Gb/s was reserved to avoid data transport on OFDM subcarriers whose electrical SNR was degraded by the presence of the IF carrier signals.It is also noted that precise synchronization was maintained between the trans-mitter-end AWGs and synthesizers.Additional details about OFDM signal generation and IQ-mixing can be found in [4],[7].To generate the optical signal,an ECL with wavelength
of1549.844nm and100kHz linewidth was employed as the CW optical source and was modulated with a12.5GHz clock signal.Subcarrier suppression was exploited to generate two optical carriers with25GHz separation[Fig.11(b)].A25GHz Gaussian optical interleaver was used for further attenuation as shown in Fig.11(c).After an EDFA,a50GHzflat optical interleaver was employed to separate the two optical carriers, with each optical carrier used to drive a20GHz IM.Each IM was modulated by an independent electrical OFDM signal with spectrum shown in Fig.11(a).The optical spectra after the modulator are shown in Fig.11(d)and(e).The two resulting optical OFDM signals were combined by a PBC,placing them on two orthogonal polarizations.The aggregate108Gb/s OFDM signal thus consisted of two polarization-multiplexed 54Gb/s signals,each containing two electrically multiplexed OFDM bands.An EDFA and a25GHzflat optical interleaver with35dB channel isolation were next used to generate thefinal SSB-POLMUX-OFDM signal shown in Fig.11(g). The resulting POLMUX-OFDM signal was next transmitted through20kmfiber and an additional15dB attenuator to the ONU receiver.Thefiber type(SMF-28)and total downstream transmitted power(8dBm)remained the same as in the40Gb/s experiment.
At the ONU receiver,the POLMUX-OFDM signal was sep-arated by a PBS and photodetected by two20GHz linear PIN photodiodes.An EDFA and tunable opticalfilter were placed before the PBS to adjust the optical power;alternately,a post-detection transimpedance amplifier may be used.The received IF OFDM signals were sampled by a Tektronix real-time os-cilloscope(DPO72004)at50Gsample/s,with all subsequent DSP done off-line.The OFDM receiver consisted of a digital IQ-demux,followed by an FFT.Both OFDM receiver outputs were fed to a MIMO PolDeMux receiver,whose operation was described in Section V.
C.Experimental Results and Discussion
Fig.12plots OSNR versus bit error rate(BER)perfor-mance for both back-to-back and20km SSMF transmission (with15dB extra attenuation)of40Gb/s and108Gb/s MIMO-OFDM-PON experiments.For reference,a54Gb/s result is also plotted in Fig.12,corresponding to single-polar-ization transmission in the108Gb/s SSB-POLMUX-OFDM system.The OSNR was measured for the optical OFDM side-band signals only without including the two optical carriers.In general,each optical carrier was approximately5–6.5dB larger
Fig.12.Experimental BER results for 40Gb/s and 108Gb/s MIMO-OFDMA-PON transmission using POLMUX-DD.
than the optical OFDM sideband.For each BER measurement
reported
here,
measured bits were evaluated.As shown by Fig.12,the best 40Gb/s MIMO-OFDM-PON BER performance was achieved after 20km transmission and
the additional attenuation
was
.Moreover,as revealed by a comparison of the back-to-back and 20km trans-mission curves,the fiber dispersion penalty is negligible.The novel training,channel estimation,and equalization methods proposed here have thus been shown to be highly effective in re-covering the 16-QAM constellation,while notably simplifying optical receiver-end complexity compared to coherent systems.For 108Gb/s transmission after 20km SSMF transmission and
the additional 15dB
attenuation,
was achieved,indicating error-free operation after FEC.Compared to the 54Gb/s curves,a 3dB penalty is noted for the POLMUX 108Gb/s result,corresponding to a doubling of the data rate at a fixed transmitted optical power.However,a comparison of the 40Gb/s and 54Gb/s BER curves at the FEC limit reveals there is no OSNR penalty for increasing the data rate over this range.Moreover,as revealed by a comparison of the back-to-back and 20km transmission curves,the fiber dispersion penalty is negli-gible in the 54Gb/s and 108Gb/s MIMO-OFDMA-PON trans-mission cases as well.
VII.C ONCLUSION
By upgrading the data rate in a fixed transmission bandwidth and enabling finely granular,dynamic bandwidth allocation,OFDM and OFDMA can significantly and cost-efficiently in-crease the data rates and flexibility of future PON-based access networks.We have proposed and experimentally demonstrated the first
single-40Gb/s and 108Gb/s MIMO-OFDMA-PON architecture for next-generation PON systems based on OFDM,POLMUX,and DD.Superior performance was exhibited after 20km SSMF transmission and a 1:32optical split.The novel POLMUX approach greatly simplified receiver-end hardware compared to coherent detectors,while increasing spectral effi-ciency to
enable
Gb/s data rates.Moreover,the proposed solution achieved the highest single-wavelength downstream transmission reported to date in any PON system.As such,the introduced architecture may be viewed as a highly attractive candidate for next-generation optical access.
R EFERENCES
[1]T.Koonen,“Trends in optical access and in building networks,”
in Proc.2008Eur.Conf.Opt.Commun.(ECOC),pp.1–31,paper We2.A.1.
[2]D.Qian,J.Yu,J.Hu,L.Zong,L.Xu,and T.Wang,“8211.5Gbps
OFDM transmission over 1000km SSMF using conventional DFB lasers and direct-detection,”IEEE Electron.Lett.,vol.44,no.3,pp.223–225,May 2008.
[3]P.Chanclou,J.-P.Lanquetin,S.Durel,F.Saliou,B.Landousies,N.
Genay,and Z.Belfqih,“Investigation into optical technologies for ac-cess evolution,”in Proc.2008IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC),pp.1–3,paper OWH1.
[4]J.Armstrong,“OFDM for optical communications,”J.Lightw.
Technol.,vol.27,no.3,pp.1–204,Jan.2009.
[5]S.L.Jansen,I.Morita,T.C.W.Schenk,and H.Tanaka,“121.9-Gb/s
PDM-OFDM transmission with 2-b/s/Hz spectral efficiency over 1000km of SSMF,”J.Lightw.Technol.,vol.27,no.3,pp.177–188,Feb.2009.
[6]A.Lowery and J.Armstrong,“Orthogonal frequency division mul-tiplexing for dispersion compensation of long-haul optical systems,”OSA Opt.Exp.,vol.14,no.6,pp.2079–2084,2006.
[7]N.Cvijetic,L.Xu,and T.Wang,“Adaptive PMD compensation
using OFDM in long-haul 10Gb/s DWDM systems,”in Proc.2007IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC),pp.1–3,paper OTuA5.
[8]I.B.Djordjevic and B.Vasic,“Orthogonal frequency division multi-plexing for high-speed optical transmission,”Opt.Exp.,vol.14,no.9,pp.3767–3775,2006.
[9]D.Qian,J.Hu,J.Yu,P.N.Ji,L.Xu,T.Wang,M.Cvijetic,and T.Ku-sano,“Experimental demonstration of a novel OFDMA-based 10Gb/s PON architecture,”in Proc.2007Eur.Conf.Opt.Commun.(ECOC),paper Mo 5.4.1.
[10]D.Qian,J.Hu,J.Yu,P.N.Ji,L.Xu,T.Wang,and M.Cvijetic,“10
Gb/s OFDMA PON for delivery of heterogeneous services,”in Proc.2008IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC),pp.1–2,paper OWH4.
[11]T.Duong,N.Genay,P.Chanclou,B.Charbonnier,A.Pizzinat,and R.
Brenot,“Experimental demonstration of 10Gbit/s upstream transmis-sion by remote modulation of 1GHz RSOA using adaptively modu-lated optical OFDM for WDM-PON single fiber architecture,”in Proc.2008Eur.Conf.Opt.Commun.(ECOC),pp.1–2,paper We.1.F.4.[12]C.T.Lin et al.,“Experimental demonstration of optical colorless di-rect-detection OFDM signals with 16-and -QAM formats beyond 15Gb/s,”in Proc.2008Eur.Conf.Opt.Commun.(ECOC),Belgium,paper Mo.3.E.1.
[13]N.Cvijetic,S.G.Wilson,and D.Qian,“System outage probability
due to PMD in high-speed optical OFDM transmission,”J.Lightw.Technol.,vol.26,no.14,pp.2118–2127,Jul.2008.
[14]B.J.C.Schmidt,A.J.Lowery,and J.Armstrong,“Experimental
demonstrations of 20Gb/s direct-detection optical OFDM and 12Gb/s with a colorless transmitter,”in Proc.2007IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC),paper PDP 18.
OFDM transmission over1000km SSMF using conventional DFB lasers and direct-detection,”in Proc.2007IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC),paper OMM3.
[16]D.Qian,J.Yu,J.Hu,P.N.Ji,and T.Wang,“11.5-Gb/s OFDM trans-
mission over0km SSMF using directly modulated laser,”in Proc.
2008Eur.Conf.Opt.Commun.(ECOC),pp.1–2,paper Mo.3.E.4. [17]W.Shieh and C.Athaudage,“Coherent optical orthogonal frequency
division multiplexing,”Electron.Lett.,vol.42,no.10,pp.587–5, 2006.
[18]A.Sano et al.,“13.4-Tb/s(1342111-Gb/s/ch)no-guard-interval co-
herent OFDM transmission over3600km of SMF with19-ps average PMD,”in Proc.2008Eur.Conf.Opt.Commun.(ECOC),pp.1–2,paper Th.3.E.1.
[19]S.L.Jansen,I.Morita,and H.Tanaka,“16252.5-Gb/s,50-GHz
spaced,POLMUX-CO-OFDM transmission over4160km of SSMF enabled by MIMO processing,”in Proc.2007Eur.Conf.Opt.
Commun.(ECOC),pp.1–2,paper PD1.3.
[20]W.Shieh,H.Bao,and Y.Tang,“Coherent optical OFDM:Theory and
design,”Opt.Exp.,vol.16,no.2,pp.841–859,2008.
[21]H.Takahashi,A.Al Amin,S.L.Jansen,I.Morita,and H.Tanaka,
“DWDM transmission with7.0-bits/s/Hz spectral efficiency using 8265.1-Gb/s coherent PDM-OFDM signals,”in Proc.2009 IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC),pp.1–3, paper PDPB7.
[22]D.Qian,N.Cvijetic,J.Hu,and T.Wang,“22.4-Gb/s OFDM transmis-
sion over1000km SSMF polarization multiplexing direct detection,”
in Proc.2009IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC), pp.1–2,paper OTuO7.
[23]D.Qian,N.Cvijetic,J.Hu,and T.Wang,“40Gb/s MIMO-OFDM
PON using polarization multiplexing and direct detection,”in Proc.
2009IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC),pp.1–2, paper OMV3.
[24]D.Qian,N.Cvijetic,J.Hu,and T.Wang,“108Gb/s OFDMA-PON
with polarization multiplexing and direct detection,”in Proc.2009 IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC),pp.1–2,paper PDPD5.
[25]W.Peng,K.Feng,and A.E.Willner,“Direct-detected polarization
division multiplexed OFDM systems with self-polarization diversity,”
in Proc.IEEE/LEOS21st Annu.Meeting,Nov.2008,pp.1–2. [26]B.J.Schmidt,Z.Zan,L.B.Du,and A.J.Lowery,“100Gbit/s transmis-
sion using single-band direct-detection optical OFDM,”in Proc.2009 IEEE/OSA Opt.Fiber Commun.Conf.(OFC/NFOEC),pp.1–3,paper PDPC3.
[27]G.J.Foschini and M.J.Gans,“On limits of wireless communications
in a fading environment when using multiple antennas,”Wireless Per-
sonal Commun.,vol.6,no.3,pp.311–335,
1998.
Dayou Qian received the B.S.degree in physics from Tsinghua University,Beijing,China,the M.S.degree in electrical engineering from the University of Cal-ifornia,Los Angeles,and the Ph.D.degree in elec-trical engineering from Florida International Univer-sity,Miami,in2000,2002,and2006,respectively. He is currently a Research Staff Member at NEC Laboratories America,Princeton,NJ.His recent re-search interests include optical/wireless systems inte-gration and novel modulation techniques for optical
transmission.
Neda Cvijetic(S’06–M’09)received the B.S.
(summa cum laude),M.S.,and Ph.D.degrees in
electrical engineering from the University of Vir-
ginia,Charlottesville,in2004,2005,and2008,
respectively.
She is currently a Research Staff Member in the
Broadband and Mobile Networking Department,
NEC Laboratories,Princeton,NJ.Her research
interests include advanced modulation/detection
techniques for high-speed optical transmission,
optical-wireless convergence,and throughput opti-mization in heterogeneous
networks.
Junqiang Hu received the B.S.degree in commu-
nication and electrical systems and the Ph.D.degree
in communication and information systems from
the University of Science and Technology of China,
Hefei,China,in1997and2002,respectively.
From2002to2005,he was a Postdoctoral
Researcher in the Department of Electrical and
Computer Engineering,University of California,
Davis.Currently,he is a Research Staff Member
at NEC Laboratories America,Princeton,NJ.His
recent research interests include router and access network architectures with an emphasis on high-performance router hardware
implementation.
Ting Wang received the M.S.degree in electrical en-
gineering from the City University of New York,New
York,and the Ph.D.degree in electrical engineering
from Nanjing University of Science and Technology,
Nanjing,China.
Since1991,he has been with NEC Laboratories
America,Princeton,NJ,where he is currently the
Department Head of optical networking research.
He has authored or coauthored more than70publi-
cations and30US patents.