Analysis of Ultrafast All-Optical OTDM Demultiplexing Based on Cascaded Wavelength Conversion in PPLN Waveguides

Abstract

Ultrafast all-optical optical time-division multiplexing demultiplexing based on the cascaded second-harmonic generation and difference-frequency generation wavelength conversion in quasi-phase-matched periodically poled lithium niobate waveguides are studied theoretically. For a typical 160- to 10-Gb/s demultiplexing process, conversion efficiency, pulse shape, time delay, phase variation, and crosstalk are investigated in two different arrangements of the 160-Gb/s signals, and both the wavelength conversion and pulse reshaping can be realized

Full text

Page 1

IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 19, NO. 7, APRIL 1, 2007495
Analysis of Ultrafast All-Optical OTDM
Demultiplexing Based on Cascaded Wavelength
Conversion in PPLN Waveguides
Yong Wang, Member, IEEE, and Chang-Qing Xu, Senior Member, IEEE
Abstract—Ultrafast all-optical optical time-division multi-
plexing demultiplexing based on the cascaded second-harmonic
generation and difference-frequency
conversion in quasi-phase-matched periodically poled lithium
niobate waveguides are studied theoretically. For a typical 160-
to 10-Gb/s demultiplexing process, conversion efficiency, pulse
shape, time delay, phase variation, and crosstalk are investigated
in two different arrangements of the 160-Gb/s signals, and both
the wavelength conversion and pulse reshaping can be realized.
generationwavelength
Index Terms—Difference frequency generation, optical time-
division multiplexing (OTDM), periodically poled lithium niobate
(PPLN), quasi-phase-matching (QPM), second-harmonic genera-
tion (SHG).
T
generation(DFG)inaquasi-phase-matched(QPM)periodically
poled lithium niobate (PPLN) waveguide has many advantages,
such as ultrafast response, low noise, high efficiency, broad
bandwidth, high dynamic range, and integration compatibility
[1]–[5]. Ultrafast all-optical demultiplexer is required in future
high-speed optical-time-division-multiplexed (OTDM) trans-
missionsystems[6].Recently,byusingthecascadedSHG-DFG
wavelength conversion technique, all-optical demultiplexing
from 40to10Gb/s [7],from 100to10Gb/s [8],and from160to
20 Gb/s [9] has been experimentally demonstrated, and some
numerical analyses have also been reported [10]–[12].
For an OTDM demultiplexing scheme based on the cascaded
SHG-DFG wavelength conversion, two input pulse trains, i.e.,
a multiplexed signal and a demultiplexing clock, are injected
into a QPM PPLN waveguide, and taken as the pump and
control waves, respectively. For example, in a typical 160- to
10-Gb/s demultiplexing case [1]–[4], the input waves comprise
a 160-Gb/s signal and a 10-Gb/s clock, and the output waves
comprise the residual 160-Gb/s signal and 10-Gb/s clock, the
residual second-harmonic (SH) wave, and the converted wave
(i.e., demultiplexed wave). The converted wave from the PPLN
waveguide has the same bit rate as the clock (10 Gb/s), carries
the code information demultiplexed from the required channel
of the 160-Gb/s signal, but has a different central wavelength.
Similar to the previous discussion in [4], there are two schemes
HE cascaded wavelength conversion based on second-
harmonic generation (SHG) and difference-frequency
Manuscript received November 3, 2006; revised December 31, 2006. This
workwassupportedbytheOntarioPhotonicsConsortium(OPC),bytheNatural
Sciences and Engineering Research Council of Canada (NSERC), and by the
Ontario Center of Excellence (OCE).
The authors are with the Department of Engineering Physics, McMaster Uni-
versity, Hamilton, ON L8S 4L7, Canada (e-mail: wangyong_x@yahoo.com).
Digital Object Identifier 10.1109/LPT.2007.893038
Fig. 1. Arrangement of pump and control waves in the two schemes. ? ?? ?
and ?
refer to the pump, control, and converted waves. (a) Scheme I.
(b) Scheme II.
to arrange 160- and 10-Gb/s pulses with respect to their wave-
lengthsandtheQPMwavelengthofdevice,asdepictedinFig.1.
In Scheme I as shown in Fig. 1(a), the 10-Gb/s clock is set at the
QPM wavelength (known as the pump wave), and the 160-Gb/s
signal is regarded as the control wave. In Scheme II as shown
in Fig. 1(b), their spectral locations are reversed in terms of the
QPMwavelength.Duetodifferentrolesofthepumpandcontrol
waves participating in this cascaded nonlinear interaction, the
convertedwaveisexpectedtobedifferentinthesetwoschemes.
Inthepreviousexperimentalandtheoreticalwork[7]–[11],only
the wavelength conversion and demultiplxing in Scheme I was
considered, while the possibility of 3R was simply discussed
by using Scheme II in [12]. In this work, from a different point
of view to the previous analyses [10]–[12], ultrafast OTDM
demultiplexing typically from 160 to 10 Gb/s are theoretically
analyzed. In particular, the conversion efficiency, pulse re-
shaping, time delay, and phase variation are comprehensively
compared between the two conversion schemes, which has not
been investigated previously to the best of our knowledge.
The device used in the simulation has similar properties to
those used in our previous experiments and simulations [4],
[5]. The poling period is 18.5
of device is 1543.0 nm. The central wavelength of the other
input wave is 1552.0 nm, and the wavelength of the converted
wave is correspondingly 1534.1 nm. The waveguide length is
typically 20 mm and allows only the TM-mode propagation.
Its normalized SHG efficiency is
power in the continuous-wave (CW) regime. Based on the de-
rived coupled-mode equations [4], the pulse propagation and
nonlinear interactions among the pump, SH, signal, and con-
verted waves in the PPLN waveguide were obtained numeri-
cally. Here, we consider nonchirped 2.0-ps Gaussian pulses for
the two inputs, and take 15-dBm injected average power and
30-dB signal-to-noise ratio in the simulation.
The 10-Gb/s converted pulses and the output optical spectra
in the two schemes are depicted in Figs. 2 and 3, and the input
m, and the QPM wavelength
190%/W at 100-mW pump
1041-1135/$25.00 © 2007 IEEE

Page 2

496IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 19, NO. 7, APRIL 1, 2007
Fig. 2. Waveforms of (a) input 160-Gb/s signal and input 10-Gb/s clock,
(b) output SH and converted waves, and (c), (d) the corresponding output
spectra in Scheme I. (a) Input wavforms. (b) Output wavforms. (c) Signal
spectrum. (d) SH spectrum.
Fig. 3. Waveforms of (a) input 160-Gb/s signal and input 10-Gb/s clock,
(b) output SH and converted waves, and (c), (d) the corresponding output
spectra in Scheme II. (a) Input wavforms. (b) Output wavforms. (c) Signal
spectrum. (d) SH spectrum.
160-Gb/s signal and the 10-Gb/s clock, as well as the SH wave
are also given. All pulses are plotted in a reference time frame
moving together with the 160-Gb/s signal pulses. In Scheme I,
as shown in Fig. 2(a), the clock pulse is synchronized to the
signalpulseinthedemultiplexedchannel.AsshowninFig.2(b),
the SH pulse is asymmetric and broader than the 10-Gb/s pump
pulseduetothegroup-velocitymismatch(GVM)[4].TheGVM
between the pump and SH pulses is 0.3 ps/mm for this device
so that the peak of the SH pulse is behind the pump pulse peak
by approximately 5 ps. The main peak of the converted pulse
is nearly symmetric and narrower than the input signal by 25%,
and has a time delay of 0.58 ps with respect to the clock pulse.
The second peak posterior to the main peak by
by the interaction between the successive signal pulse and the
delayed SH pulse in the succeeding bit. This forms a crosstalk
to the signal in the demultiplexed channel, and the crosstalk en-
ergy is about 4.3% in this case. The output spectra of the signal
and SH waves are shown in Fig. 2(c) and (d), from which we
canseethespectrumdifferencebetweentheclockandthedemul-
tiplexedsignal.SimilarlyinSchemeII,asshowninFig.3(a),the
10-Gb/sclockpulseisalsosynchronizedtothesignalpulseinthe
5 ps is caused
Fig. 4. In Scheme I, (a) converted pulse energy and crosstalk energy, (b) width
and delay of converted pulse, (c) pulse symmetry factor, and (d) maximum in-
trapulse phase variation versus clock offset. (a) Output energy. (b) Pulsewidth
and delay. (c) Symmetry. (d) Phase variation.
demultiplexed channel. In Fig. 3(b), the output SH pulse is dis-
tinctfromthatinSchemeI;theconvertedpulseisslightlyasym-
metricwithapulsewidthof1.4ps,anditspeakis0.48psposterior
totheclockpulse.UnlikeSchemeI,thecrosstalkcomponentap-
pears on the pulse leading edge. In fact, the crosstalk may come
from the interaction between the clock pulse and the two neigh-
boring SH pulses in Scheme II. The output spectra in Scheme II
are apparently different from their counterparts in Scheme I. In
particular,the converted signal spectrum is slightly broader, and
the SH spectral width is much narrower in Scheme II.
To quantitatively describe the incurred change in pulse shape
before and after the demultiplexing process, we define a vari-
able SF, namely symmetry factor, as the ratio of pulse falling
time to rising time. Here, the rising (falling) time is the dura-
tion, during which the pulse amplitude varies from 10% to 90%
of its maximum. Compared with the ideally symmetric input
Gaussian pulse SF
, the SF of the converted pulse is 1.16
and 1.25 in Schemes I and II, respectively, which means the
falling edge is longer than the leading edge in both schemes,
meanwhile, the incurred pulse distortion is more significant in
the latter scheme. One can easily understand that the pulse re-
shaping may be realized if both SF and pulsewidth can be ad-
justed appropriately. This is detailed later in this letter.
In the above analysis, the clock pulse is synchronized to the
multiplexed signal pulse. As will be shown next, an appropriate
clock offset from the signal pulse can improve the performance
of demultiplexing, and the pulse reshaping can be obtained cor-
respondingly. To quantitatively compare the conversion proper-
ties for different clock offsets in the two schemes, we focus on
such indexes as conversion efficiency, crosstalk, pulsewidth and
time delay, pulse symmetry, and intrapulse phase variation of
the converted pulse.
Varying with the clock offset, the converted pulse energy,
crosstalk energy, pulsewidth (full-width at half-maximum) and
peak delay, symmetry factor, and maximum intrapulse phase
variation with respect to the 2.0-ps Gaussian pulse are given in
Figs. 4 and 5 for Schemes I and II, respectively. We can see in
Figs. 4(a) and 5(a) that an appropriate clock offset, i.e.,
and 1.7 ps for Schemes I and II, respectively, can ensure max-
imum conversion efficiency; meanwhile, the crosstalk can be
1.5

Page 3

WANG AND XU: ANALYSIS OF ULTRAFAST ALL-OPTICAL OTDM DEMULTIPLEXING497
Fig. 5. InSchemeII,(a)convertedpulse energyandcrosstalkenergy,(b)width
and delay of converted pulse, (c) pulse symmetry factor, and (d) maximum in-
trapulse phase variation versus clock offset. (a) Output energy. (b) Pulsewidth
and delay. (c) Symmetry. (d) Phase variation.
effectively suppressed near these clock offsets. This hence im-
plies that the case of no offset is not optimal. By comparing
maximum pulse energy, we can see that Scheme I has better
conversion efficiency than the other; while Scheme II has more
reduction of noise in the converted wave. The efficiency dif-
ference between the two schemes results mainly from different
peak powers of pump pulse (different repetition rates under the
sameaveragepower)aswellasthewalk-offbetweentheSHand
signal pulses. In Figs. 4(b) and 5(b), the pulsewidth and time
delay both vary with clock offset, but have opposite trends in
the two schemes. In particular, with an increase of clock offset,
the converted pulsewidth decreases in Scheme I but increases in
Scheme II; vice versa for the time delay. This provides an effec-
tive method to adjust the demultiplexed pulsewidth in the two
schemes. From Figs. 4(c) and 5(c), one can see that the pulse
symmetry characteristic varies with clock offset. For the con-
verted pulse shape, in Scheme I, SF
edgeisalwayslongerthanitsleadingedge;inSchemeII,SFcan
be greater and less than unity. However, near the clock offset
that generates maximum conversion efficiency, the pulse sym-
metry is better in Scheme I. Together with the adjustment of
pulsewidth, an appropriate pulse reshaping can hence be real-
izedforthedemultiplexedpulseinbothschemes.Themaximum
intrapulse phase variation is shown in Figs. 4(d) and 5(d) for the
two schemes. In the whole range of clock offset, the incurred
phase change is very small, which indicates that the incurred
chirp can be neglected for the demultiplexed pulse. This is an
outstanding advantage in this kind of demultiplexing. In addi-
tion, regardless of a fixed time delay, the demultiplexed pulse
is certainly synchronized with the multiplexed and clock pulses
at the device output port, which makes further pulse manipula-
tion easier. Nevertheless, when the pulse shape is adjusted by
changing the clock offset, the output pulse energy or conver-
sion efficiency may vary. From an optimization point of view,
all these tradeoffs should be considered comprehensively.
SincethepolingperiodanddutycycleofPPLN,thewaveguide
structure as well can be engineered to a large extent [1]–[3], and
the conversion efficiency and reshaping effect of demultiplexed
pulsescanbefurtherimproved.Theserenderthiskindofdemul-
tiplexingtechniquesuitableforthefuture2Rapplications.More-
means that its falling
over,thesimulationresultsobtainedinthisworkagreewellwith
the reported experimental data [7]–[9], and further supplement
the previous theoretical analyses [10]–[12].
In conclusion, ultrafast all-optical OTDM demultiplexing
with simultaneous wavelength conversion can be realized in
QPM PPLN waveguides. When the input multiplexed signal
is taken as either the control wave or the pump wave in this
process, namely Schemes I and II, respectively, the demulti-
plexing characteristics are different. It is indicated that with the
same average power for the two inputs of the PPLN waveguide,
the conversion efficiency is higher in Scheme I than in Scheme
II; whereas the latter tends to generate narrower pulsewidth
and lower noise. By changing the clock offset with respect
to the multiplexed signal, the pulsewidth and symmetry of
the demultiplexed pulse can be appropriately adjusted, which
renders the pulse reshaping feasible. The results obtained in
this work provide some guidelines for QPM device applications
in ultrafast OTDM demultiplexing.
REFERENCES
[1] C.-Q. Xu, H. Okayama, and M. Kawahara, “1.5-?mband efficient
broadband wavelength conversion by difference frequency generation
in a periodically domain-inverted LiNbO channel waveguide,” Appl.
Phys. Lett., vol. 63, pp. 3559–3561, 1993.
[2] M. H. Chou, I. Brener, M. M. Fejer, E. E. Chaban, and S. B.
Christman, “1.55-?m-band wavelength conversion based on cascaded
second-order nonlinearity in LiNbO
Technol. Lett., vol. 11, no. 6, pp. 653–655, Jun. 1999.
[3] K.GalloandG.Assanto,“Analysisoflithiumniobateall-opticalwave-
length shifters for the third spectral window,” J. Opt. Soc. Amer. B, vol.
16, pp. 741–753, 1999.
[4] Y. Wang, J. Fonseca-Campos, C.-Q. Xu, S. Yang, E. A. Ponomarev,
and X. Bao, “Picosecond-pulse wavelength conversion based on cas-
caded second-harmonic generation-difference frequency generation in
a periodically poled lithium niobate waveguide,” Appl. Opt., vol. 45,
pp. 5391–5401, 2007.
[5] J.Fonseca-Campos, Y.Wang, B.Chen,C.-Q. Xu,S.Yang,E. A. Pono-
marev, and X. Bao, “40-GHz picosecond-pulse second-harmonic gen-
eration in an MgO-doped PPLN waveguide,” J. Lightw. Technol., vol.
24, no. 10, pp. 3698–3708, Oct. 2006.
[6] K. Uchiyama, H. Taraka, K. Mori, and T. Morioka, “160 Gbit/s
all-optical time-division demultiplexing utilizing modified mul-
tiple-output OTDM demultiplexer (MOXIC),” Electron. Lett., vol. 38,
pp. 1190–1191, 2002.
[7] C.-Q. Xu, B. Zhou, Y. L. Lam, S. Arahira, Y. Ogawa, and H. Ito,
“All-optical demultiplexing using LiNbO
wavelength converters,” Jpn. J. Appl. Phys., vol. 40, pp. L881–L883,
2001.
[8] S. Kawanishi, M. H. Chou, K. Fujiura, M. Fejer, and T. Morioka, “All-
opticalmodulation andtime-division-multiplexingof100Gbit/ssignal
using quasi-phasematched mixing in LiNbO waveguides,” Electron.
Lett., vol. 36, pp. 1568–1569, 2000.
[9] T.Ohara,H.Takara,I.Shake,K.Mori,K.Sato,S.Kawanishi,S.Mino,
T.Yamada,M.Ishii,I.Ogawa,T.Kitoh,K.Magari,M.Okamoto,R.V.
Roussev, J. R. Kurz, K. R. Parameswaran, and M. M. Fejer, “160-Gb/s
OTDM transmission using integrated all-optical MUX/DEMUX with
all-channel modulation and demultiplexing,” IEEE Photon. Technol.
Lett., vol. 16, no. 2, pp. 650–652, Feb. 2004.
[10] L. Razzari, C. Liberale, I. Cristiani, R. Tediosi, and V. Degiorgio,
“Wavelength conversion and pulse reshaping through cascaded inter-
action in an MZI configuration,” IEEE J. Quantum Electron., vol. 39,
no. 11, pp. 1486–1491, Nov. 2003.
[11] Y.Fukuchi,M.Akaike,andJ.Maeda,“Characteristicsofall-opticalul-
trafast gate switches using cascade of second-harmonic generation and
difference frequency mixing in quasi-phase-matched lithium niobate
waveguides,” IEEE J. Quantum Electron., vol. 41, no. 5, pp. 729–734,
May 2005.
[12] ——, “All-optical ultra-fast 3R repeaters using cascaded second-order
nonlinear effect in quasi-phase matched lithium niobate devices,”in 17
thAnn.MeetingIEEELasersandElectro-OpticsSociety(LEOS2004),
Puerto Rico, Nov. 7–11, 2004, pp. 977–978.
waveguides,” IEEE Photon.
quasiphase-mateched