Open Access
Add to BibSonomyAbstract
All-optical 2R regeneration of 100-Gb/s on-off-keying (OOK) signal is experimentally demonstrated based on cross gain compression (XGC) effect in semiconductor optical amplifiers (SOAs). It is shown that a high-quality logic-inverted signal and SOAs with faster gain recovery times are the two key enabling factors for obtaining regeneration results at such speeds. BER improvement of 1.2~2 dB is experimentally obtained at 1551 nm and regenerative results are demonstrated on a wide wavelength range from 1535 nm to 1555 nm. The tolerance of the input signal to optical signal-to-noise ratio (OSNR) deterioration is also experimentally studied for the 2R regeneration scheme at two different wavelengths.
© 2015 Optical Society of America
1. Introduction
The quality of optical signal can be severely degraded by amplified spontaneous emission (ASE) noise, group velocity dispersion (GVD), polarization mode dispersion (PMD)and nonlinearity accumulated in the long-haul optical fiber links, so all-optical 2R (re-amplification, re-shaping) and 3R (re-amplification, re-shaping and re-timing) regenerations are highly desirable in future all-optical networks by reducing the number of energy-consuming and speed-limiting optical-to-electrical/electrical-to-optical (OE/EO) conversions [1]. So far, all-optical signal regenerations have been demonstrated with different nonlinear medium, such as highly-nonlinearly optical fibers [2], electro-absorption modulators (EAMs) [3], semiconductor optical amplifiers (SOAs) [4] and etc. Fiber-based regenerators can be operated at very high speed, for example, regeneration of 160-Gb/s signal has been be achieved using SPM-based nonlinear optical loop mirror (NOLM) [5], however, they are usually bulky and energy-consuming. Semiconductor-based regenerators, especially those based on SOAs, featuring a lower power consumption and small footprint, are more suitable to be integrated into line-cards and terminal equipment. However, the operation speed of SOAs is limited by pattern effects due to the long carrier recovery times (several tens of picoseconds) [6]. Interferometric structures such as SOA-embedded Mach-Zehnder interferometers (MZI) [7], delay interferometers (DI) [8] and ultrafast nonlinear interferometers (UNI) [9] have been proposed and demonstrated to combat pattern effects. However, even these interferometric schemes are not immune to the pattern effect and most experimental results for SOA-based schemes are still limited to 40 Gb/s. Yoshiyasu demonstrated 84-Gb/s signal generation using a SOA-based MZI regenerator [10].
The pattern effect originates from the randomness of incoming bits which disrupts the relatively slow gain recovery of SOAs stochastically. With faster carrier recovery speeds, ultra-long SOAs with active region lengths of 4 mm were deployed for 40-Gb/s all optical regeneration [11]. Quantum-dot SOA with a fast recovery time of ~10ps was also utilized in a 80-Gb/s regeneration experiment [12]. Another effective method to combat the pattern effect is to employ the effect of cross-gain-compression (XGC) [13]. As reported recently, regeneration scheme based on XGC can be integrated into a photonic chip [14]. XGC, by simultaneously injecting two signals of different wavelengths with complementary logics into the SOA, is intrinsically immune to pattern effect when the powers of the two input signals are balanced and their sum is kept constant. Extinction ratio (ER) of both signals can be improved as a result of gain competition which always favors the strong mark signal and suppresses the weak zero. The SOA under deep saturation with almost constant optical power injection can be regarded as a high-pass filter which can effectively eliminate the amplitude noise of the input signals, especially the amplitude noise on the mark level as experiments revealed [15]. By those virtues, signal can be regenerated using XGC. Experimental results of both NRZ and RZ format have been demonstrated at 10 Gb/s and 40 Gb/s [16]. Although it is predicted that speed of regeneration by XGC can be much higher than 40 Gb/s, as long as a high-speed logic-inverted signal is available [15], solid experimental results are still required to verify such prediction. It is noted that logic-inverted wavelength conversion experiment has been demonstrated at a speed of 320 Gb/s with a SOA followed by optical filter [17], it is interesting to extend the regeneration speed by XGC into hundred-gigahertz region.
In this paper, we experimentally demonstrate for the first time 2R regeneration based on the XGC effect at a speed of 100 Gb/s with the return-to-zero on-off keying (RZ-OOK) format. To the best of our knowledge, this is the highest bitrate experimental demonstration of 2R regeneration using SOAs. It is shown that two conditions are essential to achieving regeneration at such a high speed, i.e., a high-quality logic-inverted signal and a SOA with a fast carrier recovery time. We also demonstrate that high-speed regeneration is feasible over a range of wavelengths and that this technique will only provide an improvement in signal quality within a specific region of input signal degradation.
2. Experimental setup
The experimental setup of the 2R regeneration for 100-Gb/s RZ-OOK signals is shown in Fig. 1 . 100-Gb/s RZ-OOK signal is generated with optical time-division multiplexing (OTDM) method. The 25-GHz short pulse generator has a two-stage configuration [18]. A seed pulse is firstly formed by phase modulation of continuous wave (CW) light and subsequent chirp compensation with chromatic dispersion in a segment of standard single-mode fiber (SSMF), then the seed pulse is spectrally broadened with self-phase modulation (SPM) in a highly nonlinear fiber (HNLF) and spectral-sliced with an optical band-pass filter (OBPF). By changing the wavelength of the CW laser and the center wavelength of the OBPF, 25-GHz wavelength-tunable short pulse can be obtained with a full-width at half-maximum (FWHM) pulsewidth of ~2 ps in the C-band. The 25-GHz pulse train is then intensity-modulated by a Mach-Zenhder modulator (MZM) with a 231-1 long pseudo-random bit series (PRBS). A passive 1x4 OTDM multiplexer, made of optical couplers and appropriate fiber delay-lines, is deployed to multiplex the 25-Gb/s OOK signal to 100 Gb/s. An erbium-doped amplifier (EDFA) is used to deteriorate the 100-Gb/s signal by amplified spontaneous emission (ASE) noise. An OBPF (OBPF-d) with a Gaussian shape and a 3-dB bandwidth of 5 nm is used to reject the out-of-band ASE noise. A tunable attenuator (Att) is placed in front of the EDFA to control the optical signal-to-noise ratio (OSNR) of the degraded signal.
The XGC-based 2R regenerator has a configuration modified from the one reported by Contestabile [19]. Regeneration is performed on the original signal, so that one obtains a wavelength-preserving and logic-preserving regenerated signal. The degraded 100-Gb/s OOK signal with a center wavelength of and a CW light at a different wavelength of are firstly launched into and co-propagate in an SOA (SOA1). In the SOA, logic-inverted signal is generated by cross-gain-modulation (XGM) effect. Since RZ data format is used in the experiment, the logic-inverted signal here means the signal level is completely inverted with respect to the input signal. The logic-inverted signal itself can be regarded as an inversed RZ signal. A 3-dB optical coupler is used to split the output signal of SOA into two arms. On one arm, an OBPF (OBPF1) is set to filter out the logic-inverted signal generated in SOA1 and another OBPF (OBPF2) is set on the other arm to filter out the original signal. An optical delay line (ODL) is employed to synchronize these two signals, and a tunable attenuator is used to balance the two powers. The two signals on different wavelengths with complementary logics are then combined and launched into another SOA (SOA2) where XGC comes into play for regeneration. When a logic “one” and a logic “zero” signals on different wavelengths propagate together in the SOA, they share the same saturated SOA gain and the “zero” experiences compressed gain by the co-propagating “one”, thus the ER can be improved. Besides, the SOA act as a high-pass filter for the base-band signal which suppresses the amplitude fluctuations in the mark level effectively. Finally, another OBPF (OBPF3) is used after SOA2 to filter out the regenerated signal at the original wavelength of. An EDFA was used after OBPF3 to amplify the regenerated signal for observation and measurement.
3. Experimental results and discussions
It is quite challenging to achieve the regeneration result with SOAs at the high speed of 100 Gb/s. First, there are many experimental parameters to be optimized, such as the optical powers and the wavelengths of the two signals, the driving currents of SOAs, the bandwidths and the center wavelengths of the OBPFs, and etc. Second, the degree of degradation for the original signal has to be carefully controlled because the original signal maybe further degraded by the regenerator (due to pattern effect and ASE noise additionally induced in the regenerator) if its quality is too high or too low. In this part, we firstly present a typical set of experimental conditions and the corresponding regeneration results at 100 Gb/s. We find the two most important factors for the regeneration are the high-quality logic-inverted signal and a SOA with faster carrier recovery time, as we address afterwards. The performance of the regenerator at different wavelengths and the degree of degradation to achieve a regenerative result is also investigated.
3.1 100-Gb/s 2R regeneration experiment with XGC
In the experiment, the center wavelength of the original signal was set as 1551 nm and the CW light had a wavelength of 1540 nm. The 100-Gb/s OTDM signal was degraded by ASE noise from an EDFA. The eye diagrams of the original 100-Gb/s OTDM signal captured before the Att and degraded 100-Gb/s signal captured after the OBPF-d are shown in Figs. 2(a) and 2(b) . The eye diagrams are measured with an optical sampling oscilloscope (EXFO 102-C) with a bandwidth of 500 GHz. In Figs. 2(a) and 2(b), the amplitude difference among sub-channels was induced by imbalance of the multiplexer. After degradation, the OSNR was reduced from 32 dB to 27 dB. The Q value read from the oscilloscope was decreased from 7.97 to 5.91. The average input powers of the degraded signal and the CW light to SOA1 were −2 dBm and 5.6 dBm, respectively. A commercially available SOA with long device length of 2 mm (CIP SOA-XN-C) was used for wavelength conversion by the effect of XGM. The SOA was biased at 500 mA, which provide a fiber-to-fiber small signal gain of 37 dB and a saturated output power of 15 dBm. The polarization-dependent gain (PDG) of such SOA is ~1 dB. The gain recovery time of such SOA, with appropriate arrangement of pump and probe wavelengths, can be reduced to ~10 ps due to the strong saturation by ASE [6]. The optical spectra of the signals right before and after SOA1 are shown in Fig. 2(c). The spectrum of the CW light was broadened as a result of XGM. After the SOA1, a Gaussian-shaped OBPF (OBPF2) with a center wavelength of 1551 nm and a 3-dB bandwidth of 3 nm was used to filter out the signal on the original wavelength. The eye diagram of the signal after OBPF2 is shown in Fig. 2(d). Another Gaussian OBPF (OBPF1) with a bandwidth of 1.4 nm was used to filter out the logic-inverted signal. The center wavelength of the OBPF was blue-shifted from 1540 nm with an offset of ~0.7 nm in order to obtain a logic-inverted signal with the best quality. For the XGC-based regeneration, we used another SOA with long device length of 2 mm (also CIP SOA-XN-C) which was biased at 500 mA. The input powers of the signals into the SOA were set as ~-1 dBm and ~2 dBm for 1551 nm and 1540 nm, respectively. The OBPF3 used after the SOA2 had a center wavelength of 1551 nm and a 3-dB bandwidth of 3 nm to filter out the regenerated, wavelength-preserved and logic-preserved signal. The eye-diagram of the regenerated signal after OBPF3 is shown in Fig. 2(e). By comparison of Fig. 2(b) and Fig. 2(e), it is found that the major improvement by XGC regeneration is a substantial suppression of amplitude noise on one level. The zero level noise observed in Fig. 2(b) still exists in Fig. 2(e). OTDM demultiplexing from 100 Gb/s to 25 Gb/s was performed on both the degraded signal and the regenerated signal with their bit error rate (BER) curves shown in Fig. 2(f). The receiving sensitivity at the BER of 10−9 was improved by 1.2 dB, 1.3 dB, 1.2 dB and 2 dB for all four 25-Gb/s sub-channels after regeneration, demonstrating the efficacy of the 2R regeneration scheme based on XGC even for the high-speed 100-Gb/s OOK signal. It is noticed that the amplitudes of different incoming OTDM sub-channels were different and such amplitude difference may be converted to the large difference in BER performance after nonlinear amplification in the SOA.
Fig. 2 Eye diagrams of (a) the original 100-Gb/s OOK signal and (b) the degraded 100-Gb/s OOK signal; (c) optical spectra at the input and output of SOA1; eye diagrams of signals (d) at output of SOA1 and (e) 100-Gb/s regenerated signal; (f) BER curves for the degraded and regenerated signals measured after OTDM de-multiplexing (Re. stands for regenerated signal and De. stands for degraded signal).
3.2 Key enabling factors for 100-Gb/s regeneration
Regenerative experimental results depend on many parameters and these parameters are always correlated with each other. In the experiments, we found the high quality of logic-inverted signal and the faster gain recovery time of the SOA are the most important factors.
We illustrate the importance of the quality of the logic-inverted signal by setting the center wavelength of OBPF2 at different offset from λcw. With the same wavelengths and injection powers setting as those of the previous experiment, we found that ~0.7-nm blue-shifted produces a logic-inverted signal with best signal quality. Eye diagrams of signals filtered out by OBPF with center wavelength blue-shifted 1.2nm, 0.72 nm and 0.36 nm, zero-shifted, and red-shifted 0.36 nm are shown in Figs. 3(a)-(e) , respectively. It is obviously that an OBPF with center wavelength blue-shifted 0.72nm leads to the largest eye-opening of logic-inverted signal.
Fig. 3 Eye diagrams of logic-inverted signals by detuning the OBPF at different offset with respect to the center wavelength of the CW light (a) 1.2-nm blue-shifted (b) 0.72-nm blue-shifted (c) 0.36-nm blue-shifted (d) zero shift (e) 0.36-nm red-shifted.
Eye diagrams of the regenerated signals corresponding to those logic-inverted signals are also presented in Fig. 4 . Q factors are measured from the eye diagrams to evaluate the signal quality directly in the optical domain, without involving the complexity of OE conversion and electrical filtering. The Q value of the degraded signal was 5.91 dB, and those of the regenerated signals were 3.67dB, 7.44 dB, 5.85 dB, 3.06 dB and 2.51 dB, for the different logic-inverted signals respectively. The result indicates that regeneration is possible only when the quality of logic-inverted signal is good enough. It can be understood that if the zeroes are not well suppressed in the logic-inverted signal, there will be stronger gain competition for the ones of the regenerated signal which leads to mark level fluctuations; and if the mark level in the logic-inverted signal has large fluctuation, the zero level of the regenerated signal will also be affected.
Fig. 4 Eye diagrams and Q factors of regenerative signals using logic-inverted signals filtered out by OBPF1 whose center wavelengths are (a) blue-shifted 1.2 nm (b) blue-shifted 0.72 nm (c) blue-shifted 0.36 nm (d) not shifted (e) red-shifted 0.36 nm, from the CW wavelength respectively.
Besides, we found that the polarization states of the signals at the injection of SOA1 have a strong influence on the quality of the logic-inverted signal. Two PCs were deployed at the input of SOA1 to control the polarization states of the degraded signal and the CW light. Figures 5(a) and 5(b) show the eye diagrams of the best and worst logic-inverted signals when the polarization states were changed. Although the PDG of the SOA is only 1 dB, the resultant logic-inverted signals feature with significantly different qualities. Figures 5(c) and 5(d) show the corresponding eye diagrams of the finally regenerated signal. The Q values were measured to be 3.44 and 7.35, respectively. There are another two PCs placed in front of SOA2 which also has a PDG of ~1 dB to further optimize the quality of the XGC-regenerated signal. When the quality of the logic-inverted signal was fixed, the Q value variation was measured to be ~1 dB (e.g., 1.3 dB for logic-inverted signal of the worst quality and 1.1 dB for the best one). The eye diagrams in Figs. 5(c) and 5(d) were captured when the polarization states of the PCs in front of SOA2 were optimized. In short, when SOAs with a PDG of ~1 dB are used, the PCs for logic-inverted signal generation are quite necessary because they strongly affect the overall regeneration result and the PCs for XGC are optional since they change the Q value of the regenerated signal also by ~1 dB.
Fig. 5 Eye diagrams of logic-inverted signals and regenerated signal by tuning the PCs in front of SOA1 (a) logic-inverted signal of the worst quality, (b) logic-inverted signal of the best quality, (c) regenerated signal of the worst quality, (d) regenerated signal of the best quality.
The gain recovery time of SOA for the XGC is another key factor for the performance of the regenerator. As a comparison, an SOA with shorter active region length of ~0.8mm (CIP SOA-NL-OEC-1550) was deployed for the XGC (i.e. SOA2). Bias current was set to 200mA to keep the same current density in the active region as the previous experiments. The gain recovery time for this SOA was measured to be ~90 ps. We also cascaded the 0.8-mm SOA with the 2-mm SOA for the XGC. The powers of signals launched into SOA2 and parameters of the OBPF3 were carefully optimized independently for SOA2 of different lengths (including the cascaded SOAs) in order to achieve the best performance. The regeneration result was measured with BER test as shown in Fig. 6 . The BER curves for the regenerated signal using the 0.8-mm SOA, the 2-mm SOA and the cascaded SOAs are shown in the figure. Although parameters of the regenerator were carefully optimized for the 0.8-mm SOA, signal processed by this scheme using a SOA with slow gain recovery time still had an error floor on the BER of 10−7, while signal processed using longer and cascaded SOAs achieved error-free operation and no error-floor was observed. The SOAs with longer device length (2mm-long and the cascaded ones) have faster gain recovery times because ASE noise accumulated through the length substantially deplete the gain of the SOA [6, 20 ]. According to the study of Sato, the SOA is equivalent to a high-pass baseband filter with a bandwidth inversely proportional to [21], so a longer SOA is more effective to eliminate the low-frequency fluctuation component in the input signals. The cascaded SOAs have a better regeneration performance than the 2-mm SOA. However, the sensitivity improvement at a BER of 10−9 was only 0.3 dB which is very small. We conclude that using 2-mm SOA for XGC is a good trade-off between performance and cost.
Fig. 6 BER measurements of degraded signal, regenerated signal using 0.8-mm SOA, 2-mm SOA and the cascaded 0.8-mm and 2-mm SOAs.
3.3 Wavelength-dependency of the XGC-based regeneration
In order to be compatible with the wavelength-division multiplexing system (WDM), regenerators should be able to regenerate the signals on different wavelengths. We also test the XGC-based regeneration scheme at 100 Gb/s using degraded signals at different wavelengths. 100-Gb/s OOK signals were generated and degraded with ASE noise at center wavelengths of 1535.2 nm, 1539.7 nm, 1542.0 nm, 1546.0 nm 1549.0 nm, 1551.0 nm and 1555.0 nm. The wavelength of the CW light was blue-shifted 8 nm~10 nm with respect to the center wavelength of original signal (except for the wavelength of 1535.2 nm and 1539.7 nm whose CW lights were red-shifted ~10 nm to avoid the EDFA gain dip). Experimental conditions were optimized for each wavelength to achieve the best regeneration results. Q factors of the input degraded signals and output regenerated signals are shown in Fig. 7 . Q factors of degraded signals at different wavelengths were improved by ~2 dB for almost all the wavelengths, demonstrating wavelength adaptation of the regenerator at 100-Gb/s input signal. We speculate the performance deterioration at 1535.2nm was caused by the valley of the gain spectrum of EDFA which introduced a larger noise figure to amplify the regenerated signal for observation.
Fig. 7 Q factors of degraded signals and regenerated signals and their improvements at different wavelengths from 1535 nm to 1555 nm.
It should be noted that the CW wavelengths were not optimized. In the current experiments, they were selected arbitrarily as long as an obvious regeneration result could be achieved. According to the literature [15], a better regeneration result by XGC is expected when the CW wavelength is red-shifted from that of the degraded signal by utilizing the differential gain tilt of the SOA. We tested such remark at 100 Gb/s. In the experiment, the wavelength of the degraded signal was fixed at 1550 nm, the CW wavelength was firstly set at a longer wavelength of 1560 nm and then at a shorter wavelength of 1541 nm. Other parameters were optimized respectively. Figure 8 shows the BER curves of the degraded signal, and the regenerated signal of different CW wavelength allocation. The sensitivity at a BER of 10−9 of the regenerated signal was improved by 1.7 dB when the CW wavelength was red-shifted and 2.4 dB when the CW wavelength was blue-shifted. The result is consistent with the literature, even at high speed.
3.4 Degradation tolerance
In order to investigate the tolerance to the signal deterioration degree of the regenerator, signals degraded by ASE noise of an EDFA with different degrees are processed using the regenerator at wavelength of 1546.0 nm and 1551.0 nm. Relationships between Q factors of input and output signals of the regenerator are plotted in Fig. 9 . As can be seen in Fig. 9, input degraded signals at the wavelength of 1546.0 nm has a Q-value improvement when the input Q factor is between 4.5 dB and 8.5 dB, and for signals at the wavelength of 1551.0 nm can achieve regenerative results with input Q factor between 5 dB and 8 dB. In fact, all the optical regenerators are analog devices which is capable of redistribute the noise level probability to suppress error propagation and even amplification in the subsequent transmission. As pointed out in other studies, there is a safe margin of signal degradation for an all-optical regenerator [22], if the quality of the degraded signal is too high or too low, the regenerator just further degrades the incoming signal. Our study with XGC-based regenerator at 100 Gb/s is consistent with the previous conclusions. In future applications, regenerators can be used in combination with optical performance monitoring (OPM) to activate themselves appropriately.
Fig. 9 Q factor of the regenerated signal as a function of input Q factor at wavelengths of 1546 nm (a) and 1551 nm (b), dotted-line is a reference on which output Q value is equal to input Q value.
XGC requires two OOK signals of inverted logic and reversed signal levels to maintain a constant total input optical power for the SOA. By its nature, such regeneration scheme is limited to OOK signal only. In order to extend the scheme to other advanced modulation format, appropriate format converters or decoder/encoder are still required. In our current study, only the degradation factor of ASE noise was visited and only back-to-back regeneration was performed. A systematic and complete study of the regeneration scheme under different degradation factors such as fiber dispersion, nonlinearity and timing jitter is required and the regenerator needs to be tested in a transmission experiment in the future.
3. Conclusion
In this study, we have demonstrated 2R regeneration based on the XGC effect in the SOA at a bit rate of 100 Gb/s with RZ-OOK modulation format. A 1.2-dB- 2-dB sensitivity improvement at a BER of 10−9 was achieved on the regenerated signal. We found among the many experimental parameters, it is the most important to improve the signal quality of the logic-inverted signal and to use a SOA device with faster recovery time. We also investigated the adaptation of the regeneration scheme at 100 Gb/s for different wavelengths and a regenerative result can be achieved for a wavelength range between 1535 nm~1555 nm. Tolerance to the degree of signal deterioration was also studied to find out the operation margin. The 100-Gb/s regeneration results with the working principle of XGC suggest a further potential in the speed of signal regeneration by SOA in the sub-Tb/s regime.
Acknowledgment
The authors acknowledge the support by the National Natural Science Foundation of China (No. 61275032), “973”Major State Basic Research Development Program of China (No. 2011CB301703) and Tsinghua University Scientific Research Program.
References and links
1. O. Leclerc, B. Lavigne, E. Balmefrezol, P. Brindel, L. Pierre, D. Rouvillain, and F. Seguineau, “Optical regeneration at 40 Gb/s and beyond,” J. Lightwave Technol. 21(11), 2779–2790 (2003). [CrossRef]
2. P. Mamyshev, “All-optical data regeneration based on self-phase modulation effect,” in 24th European Conference on Optical Communication (1998), pp. 475–476. [CrossRef]
3. T. Otani, T. Miyazaki, and S. Yamamoto, “40-Gb/s optical 3R regenerator using electroabsorption modulators for optical networks,” J. Lightwave Technol. 20(2), 195–200 (2002). [CrossRef]
4. D. Wolfson, A. Kloch, T. Fjelde, C. Janz, B. Dagens, and M. Renaud, “40-Gb/s all-optical wavelength conversion, regeneration, and demultiplexing in an SOA-based all-active Mach-Zehnder interferometer,” IEEE Photonics Technol. Lett. 12(3), 332–334 (2000). [CrossRef]
5. A. Bogoni, P. Ghelfi, M. Scaffardi, and L. Potì, “All-optical regeneration and demultiplexing for 160-Gb/s transmission systems using a NOLM-based three-stage scheme,” IEEE J. Sel. Top. Quantum Electron. 10(1), 192–196 (2004). [CrossRef]
6. C. S. Cleary, M. J. Power, S. Schneider, R. P. Webb, and R. J. Manning, “Fast gain recovery rates with strong wavelength dependence in a non-linear SOA,” Opt. Express 18(25), 25726–25737 (2010). [CrossRef] [PubMed]
7. T. Durhuus, C. Joergensen, B. Mikkelsen, R. J. S. Pedersen, and K. Stubkjær, “All optical wavelength conversion by SOA’s in a Mach-Zehnder configuration,” IEEE Photonics Technol. Lett. 6(1), 53–55 (1994). [CrossRef]
8. J. Leuthold, G. Raybon, Y. Su, R. Essiambre, S. Cabot, J. Jaques, and M. Kauer, “40 Gbit/s transmission and cascaded all-optical wavelength conversion over 1 000 000 km,” Electron. Lett. 38(16), 890–892 (2002). [CrossRef]
9. H. Thiele, A. Ellis, and I. Phillips, “Recirculating loop demonstration of 40 Gbit/s all-optical 3R data regeneration using a semiconductor nonlinear interferometer,” Electron. Lett. 35(3), 230–231 (1999). [CrossRef]
10. Y. Ueno, S. Nakamura, and K. Tajima, “Penalty-free error-free all-optical data pulse regeneration at 84 Gb/s by using a symmetric-Mach-Zehnder-type semiconductor regenerator,” IEEE Photonics Technol. Lett. 13(5), 469–471 (2001). [CrossRef]
11. J. Slovak, C. Bornholdt, U. Busolt, G. Bramann, C. Schmidt, H. Ehlers, H.-P. Nolting, and B. Sartorius, “Optically clocked ultra long SOAs: A novel technique for high speed 3R signal regeneration,” in Optical Fiber Communication Conference (Optical Society of America, 2004), p. WD4.
12. G. Contestabile, A. Maruta, S. Sekiguchi, K. Morito, M. Sugawara, and K. Kitayama, “Regenerative amplification in a quantum dot SOA,” in Optical Fiber Communication Conference (Optical Society of America, 2010), p. OMT2. [CrossRef]
13. G. Contestabile, R. Proietti, N. Calabretta, and E. Ciaramella, “All optical regeneration by cross gain compression in semiconductor amplifiers,” in Optical Communication,2005. ECOC 2005. 31st European Conference on(IET, 2005), pp. 415–416. [CrossRef]
14. N. Andriolli, S. Faralli, F. Bontempi, and G. Contestabile, “A wavelength-preserving photonic integrated regenerator for NRZ and RZ signals,” Opt. Express 21(18), 20649–20655 (2013). [CrossRef] [PubMed]
15. G. Contestabile, R. Proietti, N. Calabretta, and E. Ciaramella, “Cross-gain compression in semiconductor optical amplifiers,” J. Lightwave Technol. 25(3), 915–921 (2007). [CrossRef]
16. G. Contestabile, N. Calabretta, M. Presi, and E. Ciaramella, “Single and multicast wavelength conversion at 40 Gb/s by means of fast nonlinear polarization switching in an SOA,” IEEE Photonics Technol. Lett. 17(12), 2652–2654 (2005). [CrossRef]
17. Y. Liu, E. Tangdiongga, Z. Li, H. De Waardt, A. Koonen, G. Khoe, X. Shu, I. Bennion, and H. Dorren, “Error-free 320-Gb/s all-optical wavelength conversion using a single semiconductor optical amplifier,” J. Lightwave Technol. 25(1), 103–108 (2007). [CrossRef]
18. L. Huo, H. Li, D. Wang, Q. Wang, and C. Lou, “Multiwavelength 25-GHz picosecond pulse generation with phase modulation and double-side Mamyshev reshaping,” Appl. Opt. 54(18), 5703–5707 (2015). [CrossRef] [PubMed]
19. G. Contestabile, R. Proietti, M. Presi, and E. Ciaramella, “40 Gb/s wavelength preserving 2R regeneration for both RZ and NRZ signals,” in Optical Fiber Communication Conference (OSA, 2008), p. OWK1. [CrossRef]
20. G. P. Agrawal and N. A. Olsson, “Self-phase modulation and spectral broadening of optical pulses in semiconductor laser amplifiers,” IEEE J. Quantum Electron. 25(11), 2297–2306 (1989). [CrossRef]
21. K. Sato and H. Toba, “Reduction of mode partition noise by using semiconductor optical amplifiers,” in Semiconductor Laser Conference,2000. Conference Digest. 2000 IEEE 17th International (IEEE, 2000), pp. 73–74. [CrossRef]
22. L. Huo, Y. Yang, Y. Nan, C. Lou, and Y. Gao, “A study on the wavelength conversion and all-optical 3R regeneration using cross-absorption modulation in a bulk electroabsorption modulator,” J. Lightwave Technol. 24(8), 3035–3044 (2006). [CrossRef]
