All-optical 40 Gbits/s packet regeneration by means of cross-gain compression in a semiconductor optical amplifier.
ABSTRACT We experimentally demonstrate all-optical reshaping of 40 Gbits/s packets by using cross-gain compression in a semiconductor optical amplifier (SOA). This scheme, which is based on cross-saturation effects between two conjugate signals copropagating in a single SOA, is wavelength preserving and polarization independent and does not suffer from any transient effect at packet edges. We report evidence of noise redistribution and packet reshaping by means of the bit-error rate versus threshold measurements for different input optical signal-to-noise ratios.
- [Show abstract] [Hide abstract]
ABSTRACT: A review of schemes useful in a number of high-speed all-optical processing functions that exploit the gain saturation of a semiconductor optical amplifier without suffering from usual pattern-related distortions is reported. Application of pulse-limiting amplification to clock recovery and Cross Gain Compression to single and multicast conversion, 2R regeneration for NRZ, RZ and packets are discussed.01/2009;
Conference Paper: All-optical signal regeneration using SOAs[Show abstract] [Hide abstract]
ABSTRACT: This invited contribution reports recent results on all-optical signal regeneration using SOAs. Those include a scheme based on cross-gain-compression (XGC) in common SOAs for wavelength preserving 2R regeneration of RZ and NRZ signals, and regenerative amplification up to 80 Gb/s obtained thanks to detuned filtering of SPM in QD-SOA.Communications and Photonics Conference and Exhibition (ACP), 2010 Asia; 01/2011
- [Show abstract] [Hide abstract]
ABSTRACT: We introduce a novel scheme for all-optical buffering based on ar ecirculatingfiber loop with regenerative amplifi- cation by cross-gain compression in a semiconductor optical amplifier. Fifty circulations of 10-Gb/s packets with 1.2-dB power penalty every 10 rounds are reported obtaining more than 150- s buffering time. A large resilience of the scheme to noisy signals is also demonstrated.IEEE Photonics Technology Letters 01/2011; 23(22):1715-1717. · 2.04 Impact Factor
All-optical 40 Gbits/s packet regeneration by
means of cross-gain compression in a
semiconductor optical amplifier
G. Contestabile,* R. Proietti, M. Presi, S. Gupta, and E. Ciaramella
Scuola Superiore Sant’Anna, via G. Moruzzi 1, 56124, Pisa, Italy
* Corresponding author: email@example.com
Received January 11, 2008; revised April 21, 2008; accepted May 13, 2008;
posted May 22, 2008 (Doc. ID 91622); published June 24, 2008
We experimentally demonstrate all-optical reshaping of 40 Gbits/s packets by using cross-gain compression
in a semiconductor optical amplifier (SOA). This scheme, which is based on cross-saturation effects between
two conjugate signals copropagating in a single SOA, is wavelength preserving and polarization independent
and does not suffer from any transient effect at packet edges. We report evidence of noise redistribution and
packet reshaping by means of the bit-error rate versus threshold measurements for different input optical
signal-to-noise ratios. © 2008 Optical Society of America
OCIS codes: 230.1150, 250.5980.
All-optical regeneration of high bit-rate signals is a
research topic of great interest because of its poten-
tial impact on future transparent optical networks,
including packet switched ones. However, there are
only few demonstrations of packet regeneration in
the literature as, for example, in [1,2]. Indeed, when
considering all-optical packet switched networks, the
all-optical regeneration is a harder task to obtain and
to fully demonstrate. Indeed, in this case we consider
packets routed along a network in different paths
and originating from different transmitters. Hence,
they might have large packet-by-packet intensity, po-
larization, and optical signal-to-noise ratio (OSNR)
variations, as well as slight wavelength differences.
For this reason all-optical packet processing is
harder to perform. It is required, for example, to in-
clude a preprocessing stage for packet equalization
[2,3]. In addition, all the techniques must be com-
pletely polarization independent and insensitive to
wavelength differences between the packets (as is not
the case, for example, for interferometric schemes
[1,2]). Moreover, most of the schemes proposed in the
literature for all-optical regeneration provide the
wavelength conversion of the signal [1,2]; a feature
that could be undesired when exploiting WDM sys-
tems. As well, schemes employing semiconductor op-
tical amplifiers (SOAs) might suffer from transient
effects at the leading and falling edges of the packets
in correspondence with large amplifier gain varia-
tions. We recently proposed and demonstrated a
novel scheme for all-optical regeneration that works
without any wavelength shift of the signal [4,5]. This
scheme exploits cross-saturation effects between two
complementary signals (i.e., the input signal and its
inverted copy) in a saturated SOA. It is polarization
independent (once polarization-insensitive SOAs are
used), intrinsically stable, and insensitive to any in-
put packet wavelength variation. We defined this
process as cross-gain compression (XGC) and in 
we gave what was to the best of our knowledge the
first demonstration at 10 Gbits/s. In  we showed
that an improved version of the same scheme effec-
tively works at 40 Gbits/s and with both nonreturn-
to-zero (NRZ) and return-to-zero (RZ) modulation for-
mats; preliminary results at 80 Gbits/s are also
reported. In this Letter we demonstrate experimen-
tally that owing to its particular working principle,
including an assist light that saturates the SOA dur-
ing the packets’ guard times, the XGC scheme oper-
ates as well with packet–burst traffic without suffer-
ing from any distortion or transient effects at the
Here, for the sake of simplicity, we assume that
packet equalization has been implemented sepa-
rately, using an additional all-optical power equalizer
block as in [2,3], and focus on the regeneration pro-
cess, including the approximation to have all the
packets with a similar OSNR in the traffic. As will be
shown from the results in the following, packets hav-
ing different OSNR values will all experience noise
redistribution but will keep a different signal quality
at the output.
In the XGC scheme we optically generate an in-
verted signal at a different wavelength from the in-
put one and then we synchronize the two lightwaves
to copropagate in a saturated SOA, sharing the gain
compression effects. The limiting amplification in the
SOA, owing to gain saturation, and the spectral
modulation of the gain, owing to interband relax-
ations, provide noise compression on both logical “1’s”
and “0’s,” respectively . Gain saturation acts as a
power equalizer on the amplitude of 1. It works as a
high-pass filter, whose cut-off frequency is around the
inverse of the SOA’s recovery time [6,7]. Further-
more, if the interacting signals have proper wave-
length allocation, 0’s experience a negative differen-
tial gain because of the gain tilt of the amplifier [4,5].
Controlling the power of the two signals, and conse-
quently setting the condition of overall constant
power in SOA propagation, we exploit both effects
and thereby avoid pattern distortions in the SOA and
the related chirp.
As in , to generate the inverted signal at
40 Gbits/s we use the filter-assisted cross-gain modu-
lation (XGM) technique in an additional SOA .
This scheme exploits the combination of XGM and an
OPTICS LETTERS / Vol. 33, No. 13 / July 1, 2008
0146-9592/08/131470-3/$15.00© 2008 Optical Society of America
ad hoc detuned filter at the output to overcome the
usual gain recovery speed limitations. The output fil-
ter selects just part of the output spectrum from the
XGM stage, removing the slow part of the gain recov-
ery and exploiting the fast cross-phase modulation
response [8,9]. This allows for a very high-speed op-
eration for the signal inverter  and works as well
in the case of packets of any length and having any
possible guard time.
The experimental setup is reported in Fig. 1. The
input signal was generated using a 40 Gbits/s packet
transmitter at ?1=1555 nm. Packets that were
250 ns long with 200 ns guard time in between were
obtained by first using a LiNbO3 Mach–Zehnder
modulator to encode 215–1 long NRZ pseudorandom
binary sequences (PRBSs) and then an additional in-
tensity modulator to carve the packets [an example of
an eye diagram is reported in inset (a) in Fig. 1]. Note
that the scheme has been demonstrated to work for
any PRBS length [4,5] and that here the PRBS
length is determined by the packet duration. The
packets were then loaded with a variable amount of
amplified spontaneous emitted (ASE) noise from an
erbium-doped fiber amplifier (EDFA). This signal,
with around 13 dBm total power, was split in two
parts. One was sent to the XGM-based signal in-
verter. The inverter used a local tunable laser at ?2
=1565 nm with 2 dBm power, SOA1, and a bandpass
filter to generate an inverted copy of the input pack-
ets at ?2[see inset (b) in Fig. 1]. The filter was a San-
tec TF-300 with 0.7 nm 3 dB bandwidth and steep fil-
ter slopes. It was tuned by optimizing the shape and
the extinction ratio of the inverted signal. The result-
ing signal was a constant cw light in correspondence
with the guard times and with the logic inverted copy
of the data in correspondence with the input packets.
In the other arm of the regenerator the signal was
synchronized through an optical delay line with the
wavelength-converted and inverted copy. The two
waveforms were then coupled and sent to SOA2. By
controlling their power we set an almost constant
total input power to SOA2 [see inset (c) in Fig. 1].
SOA1and SOA2 (NL-OEC-1550,
?polarization-dependent gain?1 dB?
well devices with about 30 dB small signal gain,
10 dBm output saturation power at 300 mA driving
current, and around 25 ps full recovery time. The to-
tal (constant) input power at SOA2 was 3 dBm. At
the SOA2 output the reshaped packets were selected
by another bandpass filter and received. The pream-
plified receiver was made by this last bandpass filter:
an EDFA followed by another bandpass filter and an
optical attenuator to keep constant power on a
40 GHz photodiode followed by the error detector. We
performed bit-error rate (BER) gated measurements
on the packets. Following from the previous descrip-
tion, XGC occurs without any significant overall gain
modulation of SOA2, and this leads to noise redistri-
bution and waveform reshaping [see inset (d) in
An example of the appearance of input and output
packets from the regenerator on the oscilloscope is
reported in Fig. 2 for the case of packets with 25 dB
OSNR (on 0.1 nm). A clear reshaping effect on the
packet shape with no apparent distortion is evident;
moreover, owing to the copropagation of the inverted
signal, no extra noise arises in the guard time slot as,
for example, in [1,2]. To characterize and quantify the
reshaping effect we performed BER versus receiver
threshold measurements at the optimum sampling
time on the packet traffic for different noise levels of
the input signal. The BER measurements were per-
formed in gated mode, i.e., measuring the error rate
only on packets by means of a synchronization gate
signal. These kinds of measurements are the most
significant that can be performed back-to-back (with-
out recirculating loop transmission). Indeed, they ac-
count for the noise redistribution of the logical sym-
probability density functions. The results are sum-
marized in Fig. 3. In Figs. 3(a)–3(d) we report the
BER versus threshold results for input OSNR values
ranging from 25 to 40 dB together with the corre-
sponding input and output eye diagrams. We found a
large noise squeezing effect on the logical 1’s for any
OSNR value. This is due to the strong limiting am-
plification at constant power having, in this case,
around a 23 dB SOA gain compression. This effect
works with a cut-off frequency of about the inverse of
the gain recovery time: 1/25 ps=40 GHz . Regard-
ing the ground or logical 0’s level we found that noise
squeezing is mainly effective on noisier signals [see,
for example, Fig. 3(a)], while the 0 level is practically
unaffected for higher OSNR values where the noise
levels are very low. We underline, however, that the
amount of noise compression on the two logical levels
can be varied to some extent just with a slight imbal-
ance of the power of the two interacting signals. In
this case it is critical to keep the SOA gain still
largely saturated and clamped by the cw part of the
signal. This allows us to increase the noise squeezing
ser; ODL, optical delay line; SOA, semiconductor optical
amplifier; BPF, bandpass filter. Insets: (a) input signal, (b)
output from the signal inverter, (c) total input to SOA2, (d)
(Color online) Experimental setup. TL, tunable la-
and after reshaping (for 25 dB OSNR on 0.1 nm).
(Color online) 40 Gbits/s packets evolution before
July 1, 2008 / Vol. 33, No. 13 / OPTICS LETTERS
on one of the two logical levels if required. The XGC
technique squeezes the noise distributions of the logi-
cal levels and thus changes the related statistics in-
stead of increasing the signal OSNR. We observed in
all cases a significant eye opening from the oscillo-
scope traces. We note that in the case of OSNR
=25 dB, where the input eye diagram experiences a
very significant eye opening, we also found an im-
provement of the output BER value, which is in prin-
ciple theoretically unexpected if one is using opti-
mum receivers. We likely can attribute this apparent
inconsistency to the suboptimal characteristics of our
receiver (unideal optical and electrical filters, etc.).
Finally we checked that the XGC scheme prevents
all the transient effects at the packets’ leading and
falling edges by looking at magnifications of the pack-
ets’ pattern sequence at the regenerator output as re-
ported, for example, in Fig. 4. We did not find any
transients in all cases. Indeed, as can be inferred
from inset (c) of Fig. 1, the use of logically inverted
signals also prevents overall gain fluctuations in the
case of long guard times.
By using XGC between two inverted signals in a
common SOA we showed effective reshaping of
40 Gbits/s packets without any wavelength shift and
packet distortion. Owing to its working principle the
same technique could be suitable for a higher bit-rate
operation and could work as well with packets having
the RZ modulation format . Moreover, our scheme
is polarization insensitive, stable (having no inter-
ferometric element), and potentially suitable for on-
chip integration. Finally, the cascadability of the
XGC effect and its resilience to noise accumulation
(in more than 100 loops) has been very recently dem-
onstrated using a recirculating loop experiment .
This work was partially supported by the Euro-
pean FP6 program, NOBEL II.
1. M. Funabashi, Z. Pan, Z. Zhu, and S. J. B. Yoo, IEEE
Photon. Technol. Lett. 18, 1091 (2006).
2. G. T. Kanellos, D. Petrantonakis, D. Tsiokos, P.
Avramopoulos, J. Lightwave Technol. 25, 184 (2007).
3. M. Presi, N. Calabretta, G. Contestabile, and E.
Ciaramella, IEEE Photon. Technol. Lett. 19, 372
4. G. Contestabile, R. Proietti, N. Calabretta, and E.
Ciaramella, J. Lightwave Technol. 25, 915 (2007).
5. G. Contestabile,R. Proietti,
Conference, OFC 2008 (2008), paper OWK1.
6. K. Sato and H. Toba, IEEE J. Sel. Top. Quantum
Electron. 7, 328 (2001).
7. G. Contestabile, M. Presi, R. Proietti, N. Calabretta,
and E. Ciaramella, Opt. Express 15, 9849 (2007).
8. J. Yu and P. Jeppesen, J. Lightwave Technol. 19, 1316
9. Y. Liu, E. Tangdiongga, Z. Li, H. de Waardt, A. M. J.
Koonen, G. D. Khoe, X. Shu, I. Bennion, and H. J. S.
Dorren, J. Lightwave Technol. 25, 103 (2007).
10. A. D’Errico, G. Contestabile, R. Proietti, M. Presi, E.
Ciaramella, L. Bramerie, M. Gay, S. Lobo, M. Joindot,
J. C. Simon, D. Massoubre, H. Trung Nguyen, and J. L.
Oudar, in Optical Fiber Communication Conference,
OFC 2008 (2008), paper OWK4.
ments results at varying the input OSNR and correspond-
ing input–output eye diagrams for (a) 25, (b) 30, (c) 35, and
(d) 40 dB OSNR.
(Color online) BER versus threshold measure-
and falling edges at the regenerator output.
(Color online) Magnification of the packets’ leading
OPTICS LETTERS / Vol. 33, No. 13 / July 1, 2008