8th IEEE, IET International Symposium on Communication Systems, Networks and Digital Signal Processing
Reduction of Pattern-Dependent Amplitude
Modulation for RZ Data in Semiconductor
Optical Amplifier with Delay Interferometer
K.E. Zoiros*, C. L. Janer**, M.J. Connelly***, and E. Dimitriadou*
*
Department of Electrical and Computer Engineering, Democritus University of Thrace, Xanthi, Greece
de Ingeniería Electrónica, Escuela Superior de Ingenieros, Universidad de Sevilla, Seville, Spain
*** Department of Electronic and Computer Engineering, University of Limerick, Limerick, Ireland
Email: [email protected], [email protected], [email protected], [email protected]
** Departamento
Abstract—The pattern-dependent amplitude modulation in
a semiconductor optical amplifier driven by return-to-zero
(RZ) data is experimentally shown to be reduced using a
delay interferometer (DI). Pulses of this format are
amplified with far better performance than without the DI,
as quantified by the measured reduction of the amplitude
modulation from 1.6 dB to 0.29 dB and the enhancement of
the eye diagram extinction ratio from 10.5 dB to 13. 1 dB.
I.
INTRODUCTION
Semiconductor optical amplifiers (SOAs) have
technologically evolved to such extent that they have
become key elements for the development of optical
communications systems and networks. However their
exploitation in their classical amplification role is
obstructed by the pattern effect [1], which manifests as
deleterious amplitude modulation (AM). One method to
mitigate this impairment is to convert the phase change
that accompanies the irregular variation of the SOA gain
[1] to an amplitude modulation of the opposite magnitude
by means of an interferometer. For this purpose various
such schemes have been employed but limited to nonreturn-to-zero (NRZ) coding [2-4], when it is known that
in some practical applications the return-to-zero (RZ) is
preferable in terms of receiver sensitivity and fiber
transmission performance [5]. In this paper we apply this
kind of pattern effect compensation technique using a
delay interferometer (DI), as described in Sections II-III,
and experimentally manage to greatly reduce the patterndependent amplitude modulation on a 10 Gb/s RZ data
pulse stream amplified by an SOA, according to the
details presented in Section IV.
II. PRINCIPLE OF OPERATION
Fig. 1 illustrates in the form of a block diagram the
configuration under consideration which consists of a DI
serially connected to a SOA.
When the SOA is excited by a train of digitally coded
pulses whose power and duration combination saturates it
heavily, then due its finite gain recovery time the
amplified profile is not uniform. This happens because
subject to these driving conditions the gain of the SOA is
not altered in an ordinary fashion but in direct response to
the input binary content, namely it is dropped for a
logical‘1’ and partially recovered for a logical ‘0’.
Furthermore, continuous ‘0’s help the gain to rise further
978-1-4577-1473-3/12/$26.00 ©2012 IEEE
Figure 1. Block diagram of considered SOA-DI configuration
whilst continuous ‘1’s impede the gain to recover. The
result is that the gain and thus the amplitude of a specific
bit depends on the preceding bits, leading to amplitude
distortion at the output of the SOA [6]. This effect is
more severe when the logical content changes from ‘1’ to
‘0’, since a ‘1’ saturates strongly the SOA and the ‘0’ that
follows leaves time for gain recovery so that the next ‘1’
exhibits a different gain than the previous ‘1’.
The undesirable pattern dependence can be suppressed
by realizing that the irregular gain variation, which is the
cause of the problem, is accompanied by a phase change
[7]. Thus if the latter could be converted to an amplitude
modulation but of the opposite magnitude then the peaks
of the amplified marks would be balanced. This can be
achieved by exploiting the delay, ∆τ, between the two
branches of a DI formed by connecting together a pair of
50/50 couplers. A pulse that comes out from the SOA is
coupled into the input port of the DI and then it is split in
two beams of equal intensity. These are subsequently
launched into a different path and hence need different
time to travel this interferometric arrangement. This
relative temporal difference is translated to the creation of
a phase difference, namely ∆φ = 2πc∆τ/λ [8], where c is
the speed of light in vacuum and λ is the wavelength of
the injected optical signal. Thus when these beams
collide and interfere with each other at the other end of
the DI the produced result is a periodical function of the
amount of their differential phase acquired due to the
existence of ∆τ. Depending on how close this quantity is
to an even or an odd multipliciate of π, the recombined
optical power is more or less transmitted, respectively, at
the DI cross-output port. The key thus for equalizing the
peak power of the ‘1’s emerging from the SOA is to
compensate their phase change inside this active element
by the phase difference imparted between their direct and
delayed version in the DI through the proper adjustment
of ∆τ.
This concept can be also explained in the frequency
domain by recalling that the spectrum of an optical pulse
being amplified inside a SOA is broadened towards
longer wavelengths (red shift) due to the manifestation of
self-phase modulation (SPM) [7], [9]. On the other hand,
a DI acts as a frequency discriminator exhibiting at its
cross-output port a cosinusoidal transfer function whose
first null points are located at ± 1/(2∆τ) relative to the
optical carrier frequency [10]. Thus means that if these
notches are arranged through the proper adjustment of the
temporal offset of the DI to coincide with the most redshifted part of the pulse spectrum, then the latter can be
strongly attenuated and hence removed by the DI, which
is translated to a reduction of the pattern-dependent
amplitude modulation.
III. EXPERIMENTAL SETUP
Fig. 2 depicts the experimental setup. A continuous
wave (CW) beam from a 1550 nm tunable laser source is
amplified by an erbium doped fiber amplifier (EDFA)
and subsequently modulated by an electroabsorption
modulator (EAM) and a LiNbO3 Mach-Zehnder
modulator (MZM) driven by the internally synchronized
clock (CLK) and data output of a bit pattern generator
(BPG), respectively, to form a 10 Gb/s RZ 27-1
pseudorandom binary sequence (PRBS) having full-width
at half-maximum (FWHM) pulsewidth of 31 ps. It is then
amplified by an SOA, which is a 1 mm long, bulk
InGaAsP/InP device (Kamelian, model OPA-20-N-C-FA)
with a fiber-to-fiber small signal gain of 23 dB that drops
by 3 dB when the input power is -7 dBm, gain
polarization dependence of 0.5 dB, and a gain recovery
time of about 75 ps at 1550 nm, when biased at 270 mA
and thermally stabilized at 20 oC. The SOA input optical
power is controlled by an optical attenuator (VOA). The
SOA output is fed to a delay interferometer constructed
by connecting two 3 dB polarization maintaining fused
couplers with a length difference between the upper and
lower arms that results in a relative time delay, ∆τ. The
total loss inserted from this passive structure is about 7
dB. Where necessary across the whole configuration
polarization controllers (PC) are placed prior to
polarization sensitive components to ensure best coupling
of light, optical bandpass filters (OBPF) are employed to
reject the out-of-band noise and isolators (ISO) are used
to prevent undesirable back reflections.
IV. RESULTS
The average power of the pulses launched into the
SOA was -2 dBm, forcing it to operate in the heavy
saturation regime. The difference between the repetition
period of the pulses and their FWHM is less than the
SOA gain recovery time. These working conditions are
capable of provoking a pronounced pattern effect at the
SOA output. Fig. 3 shows the experimental results
obtained in the time domain using a digital
communications analyzer (DCA) with 65 GHz optical
bandwidth. For visual purposes the representative 20-bit
long segment of 11000110100101110111 contained in
the 10 Gb/s RZ 27-1 PRBS is used, which is illustrated in
the left column of Fig. 3 (a). Its logical ‘1’s have an AM
defined in [6] of 0.42 dB, while the corresponding eye
diagram in the right column has an extinction ratio (ER)
of 15.6 dB. After the SOA alone, however, these features
are impaired due to the manifestation of a strong pattern
effect, which results in the poor performance observed in
Fig. 3 (b). Indeed the marks suffer from intense amplitude
fluctuations (left column) quantified by an AM of 1.6 dB.
Moreover the eye diagram is degraded (right column) and
its ER is reduced to 10.5 dB. Nevertheless, with the use
of the DI the signal emerging from the SOA and inserted
in the structure through port 1 can be made by means of
∆τ to interfere with its delayed replica at port 2,
destructively when the incoming pulses have encountered
a partially recovered gain and constructively when they
have experienced a more saturated gain. In this manner
the pattern effect is alleviated, as shown in Fig. 3 (c),
since the peak variations of the ‘1’s are balanced (left
column) and the AM is restored to the acceptable value of
0.29 dB, while the eye diagram becomes again clear and
open having an ER of 13.1 dB and resembling that before
the SOA (right column). Notably a comparison between
Figs. 3(a) and (c) reveals that the AM of the original
pulses is also reduced from 0.42 dB to 0.29 dB. Thus the
SOA-DI combination can re-amplify and reshape the
information-carrying signal, which thus is regenerated.
This is an attractive feature when cascading many SOAs
[11], as it prevents the peak amplitude differences
between the amplified pulses from being accumulated
from stage to stage, which
otherwise would be
detrimental for the performance of an optical
transmission system.
These improvements are feasible owing to the DI
spectral response shown in Fig. 4, which has been
obtained by connecting a broadband white light source to
its input and measuring its output with an optical
spectrum analyzer (OSA) of resolution bandwidth 0.06
nm. More specifically, it exhibits a periodic comb-like
(a)
(b)
(c)
200 ps/div
Figure 2. Experimental setup
10 ps/div
Figure 3. Temporal waveforms. (a) SOA input, (b) SOA output, (c)
after DI. Left column: PRBS sample of 20 bits. Right column: Eye
diagrams
5 dB/div
(a)
(a)
1470.0
1504.0
1540.0
Wavelength (nm)
Figure 4. DI spectral response
profile having alternating maxima and minima depending
on whether the phase difference created between the two
copies of the amplified signal due to the DI relative time
delay is an even or odd multiple of π, respectively. The
wavelength spacing between adjacent peaks or free
spectral range (FSR) is approximately 5.4 nm, which
from [8] FSR = (λ2)/(c∆τ), where λ = 1550 nm, means
that the value of the employed time delay is ∆τ = 1.48 ps.
This form of the transfer function allows to exploit the DI
as a notch filter and suppress the spectral components of
the amplified pulses that have been spread towards the
longer sideband [9], as shown in Fig. 5(b) compared to
Fig. 5(a) before the SOA. This can be done by biasing
through ∆τ the DI at the quadrature point with negative
slope versus the optical carrier wavelength in the
transmission characteristic [12]. In this manner these
spectral components are forced to lie close to the null
points, which are located at the middle of the FSR and
have a maximum relative attenuation of 14 dB.
Consequently the DI eliminates the most red-shifted part
of the pulse spectrum, as shown in Fig. 5(c), which is
associated with the pattern-dependent distortion caused
by strong gain saturation [9]. On the other hand the
performance of the delayed-interference configuration
was very sensitive to changes in the environmental
conditions so that it was not possible to make bit error
rate measurements since error-free operation was
achieved only for a very short time slot. Nevertheless this
is not a fundamental problem but rather a technical
difficulty that can be overcome if the phase difference
incurred between the DI arms is controlled by active
means, such as a temperature controller [4] or a feedback
circuit [12].The fabrication of the DI using Si-SiO2
waveguide technology [8] or its monolithic integration
with the SOA in a fully packaged compact module [13]
can enhance further the stability of the scheme.
V. CONCLUSION
The feasibility of employing a DI to reduce the patterndependent amplitude modulation induced on RZ data
when amplified by an SOA has been experimentally
demonstrated. The proposed scheme has enabled to
realize significant performance improvements compared
to the SOA alone, and has helped drop the AM to a low
level even below that being present on the original data
before the SOA. This suggests that it can efficiently
compensate the pattern effect and its negative impact on
this pulse format.
(b)
(b)
(c)
(c)
Figure 5. Optical spectra. (a) SOA input, (b) SOA output, (c) after DI
REFERENCES
[1]
[2]
[3]
[4]
[5]
M.J. Connelly, Semiconductor Optical Amplifiers. Dordrecht, The
Netherlands: Kluwer Academic Publishers, 2002.
Q. Xu, M. Yao, Y. Dong, W. Cai, and J. Zhang, “Experimental
demonstration of pattern effect compensation using an
asymmetrical Mach-Zehnder interferometer with SOAs,” IEEE
Photon. Technol. Lett., vol. 13, pp. 1325-1327, December 2001.
K. Chan, C.-K. Chan, W. Hung, F. Tong, and L.K. Chen,
“Waveform restoration in semiconductor optical amplifier using
fiber loop mirror,” IEEE Photon. Technol. Lett., vol. 14, pp. 995997, July 2002.
C.S. Wong, and H.K. Tsang, “Reduction of bit-pattern dependent
errors from a semiconductor optical amplifier using an optical
delay interferometer,” Opt. Commun., vol. 232, pp. 245-249,
March 2004.
P.J. Winzer, and R.-J. Essiambre, “Advanced modulation formats
for high-capacity optical transport networks,” J. Lightwave
Technol., vol. 24, pp. 4711-4728, December 2006.
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
K.E. Zoiros, T. Siarkos, and C.S. Koukourlis, “Theoretical
analysis of pattern effect suppression in semiconductor optical
amplifier utilizing optical delay interferometer,” Opt. Commun.,
vol. 281, pp. 3648-3657, July 2008.
G.P. Agrawal, and N.A. Olsson, “Self-phase modulation and
spectral broadening of optical pulses in semiconductor laser
amplifiers,” IEEE J. Quantum Electron., vol. 25, pp. 2297-2306,
November 1989.
H. Dong et al., “Multiwavelength fiber ring laser source based on
a delayed interferometer,” IEEE Photon. Technol. Lett., vol. 17,
pp. 303-305, February 2005.
K. Inoue, “Optical filtering technique to suppress waveform
distortion induced in a gain-saturated semiconductor optical
amplifier,” Electron. Lett., vol. 33, pp. 885-886, May 1997.
X. Tang, N.Y. Kim, and J.C. Cartledge “Noise transfer
characteristics in a semiconductor optical amplifier with
application to wavelength conversion based on a delay
interferometer,” J. Lightwave Technol., vol. 26, pp. 1715-1721
June 2008.
S. Singh, and R.S. Kaler, “Transmission performance of 20 x 10
Gb/s WDM signals using cascaded optimized SOAs with OOK
and DPSK modulation formats,” Opt. Commun., vol. 266, pp. 100110, October 2006.
D. Derickson, Fiber optic test and measurement. Prentice Hall
Professional Technical Reference, 1997, Chap. 5.
J. Leuthold et al., “100 Gbit/s all-optical wavelength conversion
with integrated SOA delayed-interference configuration,”
Electron. Lett., vol. 36, pp. 1129-1130, June 2000.