Content uploaded by James Roy Taylor
Author content
All content in this area was uploaded by James Roy Taylor on Oct 02, 2014
Content may be subject to copyright.
Tunable repetition-rate multiplication of a 10 GHz pulse train
using linear and nonlinear fiber propagation
C. J. S. de Matosa) and J. R. Taylor
Femtosecond Optics Group, Imperial College, Prince Consort Road, London, SW7 2BW United Kingdom
共Received 12 August 2003; accepted 4 November 2003兲
The temporal Talbot effect and soliton propagation in an optical fiber were exploited to yield a series
of pulse train sources with tunable repetition rate simply through variation of the pulse train power
in sections of the fiber. In a dual-repetition-rate configuration, 10 and 20 GHz or 10 and 30 GHz
repetition rates could be achieved depending on the fiber length used, with pulse durations lower
than 21 ps. In a triple-repetition-rate configuration, 10, 20, and 30 GHz repetition rates were
obtained, with pulse durations lower than 15 ps. © 2003 American Institute of Physics.
关DOI: 10.1063/1.1636824兴
High repetition rate pulse sources are essential for up-
grading current optical telecommunication systems and have
received significant attention in the past few years. Sources
of up to ⬃40 GHz repetition rate can be directly obtained
electronically with the use of Mach–Zehnder electro-optic
modulators1and electroabsorption modulators.2Optical tech-
niques, such as modulational instability,3rational
mode-locking,4split semiconductor laser cavities,5and laser
beating,6offer more repetition rate flexibility and can be used
to achieve higher frequencies.
All-optical repetition rate multiplication through the
temporal Talbot effect has been recently proposed as means
to obtain high-frequency pulse sources.7,8 In this simple and
flexible technique, a pulse train at a certain repetition rate is
chromatically dispersed in an optical fiber7or in a chirped
fiber Bragg grating8to the point that neighboring pulses
overlap and interfere with one another. By carefully adjust-
ing the fiber or grating dispersion, the interference can be
such that a multiple of the original repetition rate is obtained.
The use of an optical fiber as the dispersive element is par-
ticularly interesting, as the Talbot effect achieved by linearly
propagating the pulse train in a section of the fiber can be
employed in combination with nonlinear soliton propagation
of the train in a different section to result in useful and flex-
ible devices. Recently, 4⫻repetition rate multiplication and
Raman-assisted temporal soliton compression in the same
fiber have been demonstrated.9
In this letter we present a pulse source with selectable
repetition rate based on the propagation of a 10 GHz pulse
train in an optical fiber. By varying the power launched into
the fiber it was possible to either obtain repetition rate mul-
tiplication due to the temporal Talbot effect or to maintain
the original repetition rate through soliton transmission. Us-
ing this concept, a repetition rate selectable between 10 and
30 GHz or between 10 and 20 GHz was obtained using two
different lengths of fiber. By modifying the configuration it
was possible to obtain soliton propagation or repetition rate
multiplication independently in two different fiber sections
leading to 10, 20, and 30 GHz selectable repetition rate from
a single device.
The experimental setups used for the repetition-rate-
selectable pulse sources are shown in Fig. 1. Figure 1共a兲
shows the first setup used, which allows selectivity between
two repetition rates. Pulses at 1549.4 nm with durations of
⬃6.8 ps were generated in a 10 GHz pulse source and am-
plified in an erbium-doped fiber amplifier 共EDFA兲before be-
ing launched into a length of standard telecommunication
fiber 共STF兲. The pulse source consisted of a cw, single-
longitudinal-mode semiconductor laser, an electroabsorption
modulator, and a linearly chirped fiber Bragg grating used to
compensate for the chirp produced by the modulator. The
gain of the EDFA was variable and was adjusted to give
in-STF average powers that were either above or below the
fiber soliton threshold for the input pulses used 共estimated to
be ⬃100 mW). STF lengths of 25.4 and 38.3 km were sepa-
rately used and, when the pulse train power was below the
soliton threshold, gave repetition rate multiplication factors
of 3 and 2, respectively. When the pulse train power met the
soliton threshold, a 10 GHz output train was obtained as
expected. The output pulse train was simultaneously ana-
lyzed in an optical spectrum analyzer and in an autocorrela-
tor. A 20 ps impulse response detector and a 50 GHz digital
sampling oscilloscope were also used to directly probe the
pulse train temporally but in this case the actual pulse shape
was hindered by the detector resolution.
a兲Electronic mail: c.de-matos@ic.ac.uk
FIG. 1. Experimental configurations for the repetition-rate-selectable pulse
source; 共a兲setup for obtaining 10 and 20 GHz or 10 and 30 GHz repetition
rates; 共b兲setup for obtaining 10, 20, and 30 GHz repetition rates.
APPLIED PHYSICS LETTERS VOLUME 83, NUMBER 26 29 DECEMBER 2003
53560003-6951/2003/83(26)/5356/3/$20.00 © 2003 American Institute of Physics
Downloaded 27 Feb 2006 to 155.198.207.26. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
Figure 1共b兲shows the setup used for obtaining 10, 20,
and 30 GHz selectable repetition rate from a single device. In
this case, 38.3 km of STF were split into two sections of 12.9
and 25.4 km so that either linear or nonlinear propagation
could be independently obtained in each section. The same
10 GHz pulse source was utilized and the power in each fiber
section was controlled by adjusting the attenuation 共Atten兲
before each section. When the pulse train power was below
the soliton threshold in both fiber sections, the temporal Tal-
bot effect led to an output repetition rate of 20 GHz. When
the train power met the soliton threshold only in the 12.9 km
STF, temporal Talbot effect in the 25.4 km STF generated a
30 GHz output repetition rate. Finally, when soliton propa-
gation occurred in both fibers, a 10 GHz output pulse train
was obtained.
It was found that when the pulse train propagated in the
soliton regime in the 12.9 km STF, pulse energy shedding
occurred both due to nonadiabatic soliton attenuation in the
STF 共the STF attenuation was measured to be 0.19 dB/km兲
and due to the fact that the pulses generated by the 10 GHz
pulse source had a small chirp. This energy shedding led to
pulse train spectral degradation that prevented quality repeti-
tion rate tripling to be obtained in the subsequent, 25.4 km
STF section. To improve the input pulse train quality, a non-
linear optical loop mirror 共NOLM兲was added between the
EDFA and the 12.9 km STF, which comprised a 32:68 cou-
pler, a ⬃2 km STF, and a polarization controller 共PC兲. The
NOLM operated in the nonlinear switching regime, output-
ting solitons. The EDFA output power and the NOLM polar-
ization were adjusted while the configuration was set to 30
GHz operation so that the quality of the 30 GHz output pulse
train was optimized. These parameters were then left con-
stant for the 10 and the 20 GHz repetition rate cases. To
overcome the soliton attenuation, Raman amplification was
provided in the 12.9 km fiber section by a counterpropagat-
ing Raman pump consisting of a fiber Raman laser 共FRL兲at
1455 nm. Optical circulators OC2 and OC1 were, respec-
tively, used to insert and extract the Raman pump to and
from this fiber section. The FRL output power was also cho-
sen in order to optimize the 30 GHz repetition rate case and
was increased only in the 10 GHz case, in which the power
reaching the 25.4 km STF had to meet the fiber soliton
threshold. The output pulse train was analyzed as in the case
of the previous setup. Alternatively, both sources of soliton
energy shedding could be addressed by placing the NOLM
between fiber sections so as to spectrally clean the 25.4 km
STF input pulses. This scheme would require an additional
EDFA and was not tested in the present work.
Figure 2 depicts the autocorrelation traces of the output
pulse train when the configuration illustrated in Fig. 1共a兲was
used. Figure 2共a兲shows the case when the 38.3 km STF was
employed and the in-STF average pulse train power was set
to ⬃130 共top兲and ⬃18 mW 共bottom兲. When the train power
met the soliton threshold, a 10 GHz output pulse train with
pulse durations of ⬃21 ps was obtained. The pulse broaden-
ing observed can be accounted for by the fiber attenuation
that reduces the soliton peak power and consequently leads
to an increase in soliton duration. When the input pulse train
power was below the fiber soliton threshold, the temporal
Talbot effect led to a 20 GHz output pulse train with 8.5 ps
pulses. The slight pulse duration increase in this case is due
to a small deviation from the optimal fiber length required
for repetition rate doubling. Figure 2共b兲shows similar traces
for when the 25.4 km STF was employed. The top trace
shows the autocorrelation of the output pulse train for an
average input power of ⬃140 mW. As in the previous case,
soliton propagation prevented the temporal Talbot effect to
take place and a 10 GHz pulse train with pulse durations of
⬃15 ps was obtained. The bottom trace shows the autocor-
relation of the output pulse train for an average input power
of ⬃11 mW. In this case the temporal Talbot effect tripled
the repetition rate and a 30 GHz train of 7.2 ps pulses was
achieved. For both fiber lengths used the temporal traces
obtained with the sampling oscilloscope and the detector in-
dicated that the pulse trains had a low amplitude fluctuation
of ⬃8% or less.
Figure 3 shows autocorrelations of the output pulse train
at the three different repetition rates obtained with the ex-
perimental configuration depicted in Fig. 1共b兲. For all cases
shown, the pulse train at the NOLM output had an average
power of 120 mW and a pulse duration of 9.7 ps. The 10
GHz output train of 14.5 ps pulses obtained when no attenu-
ation was used can be seen in Fig. 3共a兲. The FRL power
utilized was 590 mW, providing an internal gain 共ratio be-
tween Raman gain and fiber loss兲of 3.6 dB and inducing
soliton temporal compression by a factor 2.1. Figure 3共b兲
illustrates the case in which 18 dB attenuation was induced
FIG. 2. Autocorrelation traces of the output pulse train obtained with the
configuration shown in Fig. 1共a兲for input powers above 共top兲and below
共bottom兲the soliton threshold, and for STF lengths of 38.3 共a兲and
25.4 km 共b兲.
5357Appl. Phys. Lett., Vol. 83, No. 26, 29 December 2003 C. J. S. de Matos and J. R. Taylor
Downloaded 27 Feb 2006 to 155.198.207.26. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
prior to the 12.9 km STF and the FRL power was set to 390
mW. Linear propagation along both fiber sections resulted in
a 20 GHz train of 10 ps pulses. The case in which 18 dB
attenuation was induced before the 25.4 km STF can be seen
in Fig. 3共c兲. The 390 mW FRL provided an internal gain of
1.8 dB and soliton compression by a factor 1.9. Repetition
rate multiplication in the 25.4 km fiber section resulted in a
30 GHz train of 6.1 ps pulses. The pedestal levels observed
in Figs. 2 and 3 are believed to result from two main sources.
In conditions involving repetition rate multiplication, the
presence of spurious spectral components and deviations
from the optimum fiber length and dispersion are expected to
have increased the pedestal. In conditions involving nonlin-
ear propagation, the soliton energy shedding processes men-
tioned earlier are believed to have led to some pedestal, even
with the use of the NOLM. Further optimizing the fiber pa-
rameters, the input pulse quality, and the Raman gain distri-
bution along the fiber could improve the pulse-to-pedestal
ratio.
When the pulse train generated by the configuration in
Fig. 1共b兲was measured using the sampling oscilloscope and
the 20 ps detector ⬃17% and ⬃16% amplitude fluctuations
were observed in the 20 and 30 GHz cases, respectively. A
simulation of the temporal Talbot effect indicated that these
fluctuations were a consequence of the imperfect pulse train
spectral profile obtained at the NOLM output, in the 20 GHz
case, and at the 12.9 km STF output, in the 30 GHz case. The
simulation also showed that a significant reduction on such
fluctuations could be obtained if the input pulse train quality
is improved.
Due to the different nature of light propagation in the
linear and nonlinear regime, one could expect to observe
different polarization states when solitons or repetition-rate-
multiplied pulses reached the autocorrelator. In this work,
this feature was not observed.
In conclusion, fiber pulse train sources have been dem-
onstrated that offer selectable repetition rate. Depending on
the configuration used, 10 and 20 GHz, 10 and 30 GHz, or
10, 20, and 30 GHz repetition rates could be obtained simply
by varying the pulse train power. Pulses with durations equal
to or lower than 21 ps were achieved. The dual-repetition-
rate configurations exhibited low pulse train amplitude fluc-
tuations. ⬃17% amplitude fluctuations were observed in the
triple-repetition-rate configuration but could be reduced if
higher-quality input pulse trains are used. The technique de-
scribed here can be readily applied to obtain other repetition
rates simply by choice of the fiber length used.
C.J.S.d.M. is supported by Coordenac¸a
˜
o de Aperfeic¸oa-
mento de Pessoal de Nı
´vel Superior 共CAPES兲—Brazil and
an Overseas Research Student 共ORS兲award—UK.
1Th. Pfeiffer and G. Veith, Electron. Lett. 29, 1849 共1993兲.
2D. Moodie, A. Ellis, X. Chen, F. Garcia, D. Rogers, S. Perrin, P. Cannard,
R. McLaughlin, M. Robertson, S. Amos, S. Cole, C. Ford, and I. Reid,
OSA Trends in Opt. Photonics Ser. 86, 267 共2003兲.
3M. Nakazawa, K. Suzuki, and H. A. Haus, IEEE J. Quantum Electron. 25,
2036 共1989兲.
4E. Yoshida and M. Nakazawa, Electron. Lett. 32, 1370 共1996兲.
5Y. K. Chen, M. C. Wu, T. Tanbun-Ek, R. A. Logan, and M. A. Chin,Appl.
Phys. Lett. 58, 1253 共1991兲.
6S. V. Chernikov, J. R. Taylor, P. V. Mamyshev, and E. M. Dianov, Elec-
tron. Lett. 29, 931 共1992兲.
7I. Shake, H. Takara, S. Kawanishi, and M. Saruwatari, Electron. Lett. 34,
792 共1998兲.
8J. Azan
˜
a and M. A. Muriel, Opt. Lett. 24, 1672 共1999兲.
9D. A. Chestnut, C. J. S. de Matos, and J. R. Taylor, Opt. Lett. 27, 1262
共2002兲.
FIG. 3. Autocorrelations of the output pulse train obtained with the configu-
ration shown in Fig. 1共b兲for when nonlinear propagation occurred in both
fiber sections 共a兲, for when linear propagation occurred in both fiber sections
共b兲, and for when nonlinear propagation occurred in the 12.9 km STF and
linear propagation occurred in the 25.4 km STF 共c兲.
5358 Appl. Phys. Lett., Vol. 83, No. 26, 29 December 2003 C. J. S. de Matos and J. R. Taylor
Downloaded 27 Feb 2006 to 155.198.207.26. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp