A simple all-optical power equalization scheme based on a single two-section reflective semiconductor optical amplifier (RSOA) is presented. Double optical path and non-uniform injection current density in the two sections easily saturate the RSOA and suppress pattern effect, thereby significantly reducing packet-to-packet power fluctuation while maintaining better signal quality. The mechanism of the two-section RSOA-based power equalizer is investigated and it is indicated that the two-section RSOA biased at proper current density functions as three cascaded SOAs, including a preamplifying SOA, a gain-saturated SOA and a third SOA. The performance dependence on driven current and structural parameters is also studied.
© 2013 OSA
The increasing user bandwidth demands have been driving the rapid growths of passive optical networks (PONs) in recent years. Data packets from different optical network units (ONUs) pass through different optical paths and experience unequal optical losses, therefore, they may exhibit large variations in optical power on reaching the receiver end at optical line terminal (OLT), which causes serious problems at the receiver with a limited dynamic range. To cope with all possible incoming power levels, one has to develop a fast and accurate burst-mode receiver with a large input power dynamic range, which brings about excessive complexity and high cost to the receiver end. If optical power equalization technique is employed before detection process, then the stringent requirements on the receiver end can be relaxed, and the burst-mode receiver can become less complex with less demanding dynamic range, or even inexpensive continuous-mode receiver can still be suitable.
Up to now, optical power equalization schemes using erbium-doped fiber amplifiers , semiconductor optical amplifiers (SOAs) [2, 3] have been demonstrated. One kind of SOA-based all-optical equalizer is to dynamically adjust the gain of a SOA or a gain-clamped SOA and it needs complex opto-electronic circuit controls [4, 5]. Another kind of SOA-based all-optical equalizer is based on self-gain modulation caused by gain saturation, which offers smaller optical gain to higher power level packets, thereby reducing packet-to-packet power variation. In order to make a SOA saturate, input optical power should be high enough, and at least two cascaded traveling-wave SOAs (TW-SOAs) are usually needed, with the first SOA acting as a preamplifier to drive the second SOA into saturation regime . However, gain saturation inevitably accompanies pattern effect and waveform distortion, which deteriorates signal quality. In order to improve signal quality, one had to utilize a gain-saturated SOA together with an assisting light [2, 3], or combined multiple SOAs with interferometric arrangement [3, 7].
An optical power equalizer based on a two-section TW-SOA was also demonstrated . In order to obtain constant output power with good signal quality, the injection current in the first section was adjusted for various input power to offer constant input power of the second section . So, it still required electronic circuit controls. In this paper, we propose a novel and simple all-optical equalization scheme based on a single two-section reflective optical amplifier (RSOA). Double optical path and non-uniformly injected current density in the two sections are beneficial to reducing packet-to-packet power variation, suppressing pattern effect and improving signal quality. And, in comparison with other schemes, a two-section RSOA-based all-optical equalizer has advantages like simpler configuration and less complexity since it just requires a single RSOA and an optical circulator to retrieve the output signal, thus avoiding the use of multiple SOAs, other expensive components or complex electronic circuit controls.
The organization of the paper is as follows, section 2 presents the principle, section 3 gives the results and discussion, and finally section 4 draws a conclusion.
The schematic diagram of a two-section RSOA is shown in Fig. 1 . There is a narrow insulation channel of length L3 to separate the RSOA into two sections of length L1 and L2, respectively. The two sections possess identical active region structure and other internal structures, but they have their own independent current sources, I1 and I2. The current density ratio is defined as J1/J2, where J1 = I1/ (WL1) and J2 = I2/ (WL2), in which W is the width of the active region. Optical signals enter the RSOA from the front facet, which has an anti-reflective coating to ensure very low reflectivity R1, allowing more signals to enter the RSOA. The rear facet has a high-reflection coating for larger reflectivity R2, hence effectively reflecting the incoming optical signals. Finally, the optical signals output from the front facet.
In a TW-SOA, an optical signal travels single path in its cavity because its front facet and rear facet both have near-to-zero reflectivity. In contrast, the optical signal travels double paths in a RSOA and gets larger optical gain thanks to the high reflectivity of the rear facet. This also means more carriers are depleted; as a result, a RSOA is more likely to experience gain saturation than its TW-SOA counterpart. Therefore, just a single RSOA is potentially sufficient for all-optical power equalization, even though the input power is relatively low; and there is no need of using another SOA to preamplify the input optical signal.
Nevertheless, when common TW-SOAs or common RSOAs operate in saturation regime, the extinction ratio (ER) of the optical signal degrades as a result of gain saturation, and Q factor decreases due to pattern effect and the reduction of ER. As we know, common TW-SOAs or common RSOAs have only one section and one electrode, current is uniformly injected into their cavities; by contrast, a two-section RSOA with two independent electrodes enables the adjustment of carrier density distribution by changing J1/J2. Since gain saturation and pattern effect strongly rely on carrier density and its distribution in the cavity, it is possible to mitigate gain saturation and pattern effect when the carrier density distribution is optimized by choosing appropriate J1/J2. To further clarify these, we analyze the signal evolution process in a two-section RSOA. It can be classified into 5 stages, as schematically shown in Fig. 2 . Stage 1 and stage 2 relate to the forward propagation in the first section and the insulation channel, respectively. Stage 3 includes the signal propagation in the second section, that is, the optical pulse travels forward at first and then is reflected by the rear facet and propagates backward in the 2nd section again. Stage 4 and stage 5 refer to the backward propagation in the insulation channel and 1st section, respectively.
The optical pulse evolution process in each stage is simulated by a segmentation model which considers the longitudinal distribution of carrier density in the RSOA cavity [9, 10]. In addition, carrier diffusion effect, amplified spontaneous emission (ASE) and gain compression effect are also taken into account [10, 11]. The active region of the MQW-RSOA is divided into multiple subsections along the optical propagation direction (i.e., z axis), and carrier rate equation in subsection i is [9, 10]:10]. ,,andare the photon density of forward-traveling, backward-traveling optical signal and the ASE, respectively. D is diffusion coefficient, and R (Ni) is given by11]:12, 13]:Table 1 .
Figures 3 (a1)-3(e1) and Figs. 3(a2)-3(e2) depict the normalized time-domain waveform evolution of the optical signal in each stage when the input power Pin is −30 dBm. The RSOA is biased at J1/J2 of 1 (i.e., I1 = 40 mA, I2 = 360mA) and 4 (i.e., I1 = 123 mA, I2 = 277 mA), respectively; and the optical packets are 10 Gb/s nonreturn-to-zero (NRZ) data with 27-1 pseudo-random binary sequence. The lengths of the first section, the insulation channel and the second section are 120, 20 and 1060 μm, respectively. And for each stage depicted in Fig. 2, the input pulse and the output pulse are given by dashed line and solid line, respectively. It should be noted that the waveforms of the input pulse and output pulse are normalized by their maximal power levels.
It can be seen from Figs. 3(a1) and 3(a2) that the optical pulse suffers from slight pattern effect during the amplifying process at stage 1. In this sense, section 1 acts as a preamplifying SOA. It is also noted that pattern effect at J1/J2 of 4 is less obvious than that at J1/J2 of 1. This is because when J1/J2 is 4, section 1 has more carriers to amplify the optical signal.
When the optical pulse travels in the insulation channel, it experiences optical absorption since there is no current injection. In this sense, the insulation channel functions as a saturable absorber and helps to mitigate gain saturation , as can be seen from Figs. 3(b1) and 3(b2), Figs. 3(d1) and 3(d2). Although the signal suffers from power loss during stage 2 and stage 4, the loss is around 1 dB because the length of the insulation channel is very small. And it should be noted that the loss cannot be deduced from Figs. 3(b1), 3(b2), 3(d1) and 3(d2), because the optical power is normalized.
When the optical pulse passes through section 2 biased at J2, serious pattern effect takes place because no sufficient carriers are supplied to the optical pulse and gain saturation occurs. In this sense, the 2nd section biased at J2 is equivalent to a gain-saturated SOA. It can be seen from Figs. 3(c1) and 3(c2) that pattern effect is less distinct and the degradation of extinction ratio is also smaller when the RSOA is biased at J1/J2 of 4. This is because when J1/J2 is 4, less current density in section 2 helps to retard the rapid growth of the signal photons and the occurrence of gain saturation.
After passing backward through the insulation channel, the optical pulse enters section 1 again and gets amplified at stage 5. It can be seen from Figs. 3(e1) and 3(e2) that when J1/J2 is 4, waveform distortion at ‘1’ bits gets compensated to some degree and the extinction ratio is slightly improved because 1st section with higher injection current density can provide more carriers. By contrast, the optical signals see a saturated SOA with slight pattern effect especially at ‘0’ bits when J1/J2 is 1.
In brief, a two-section RSOA biased at proper J1/J2 is equivalent to three cascaded SOAs to some extent: the first SOA pre-amplifies the optical signal, the second SOA works as a saturated SOA, and the third SOA further amplifies the optical signals. Therefore, gain saturation with less pattern effect can be obtained by optimizing the current distribution in the two sections, and smaller power variation can be achieved at less cost of signal quality.
3. Results and discussion
We simulate a system shown in Fig. 4 to investigate the performance of a two-section RSOA-based all-optical power equalizer by using software tool Optiwave-Optisystem. An optical transmitter (Tx) at 1.3 μm wavelength generates optical packets of 10 Gb/s nonreturn-to-zero (NRZ) data with 27-1 pseudo-random binary sequence and extinction ratio of 10 dB, a variable optical attenuator (VOA) is used to tune the packet power. After passing through a 20-km single-mode fiber (SMF) without dispersion compensation, the packets are launched into a 1.3 μm multiple-quantum well (MQW) two-section RSOA via an optical circulator (OC). The reflected optical packets from the RSOA are sent into the OC again and detected by a PIN photodetector-based receiver (Rx).
The average optical power of a packet before the RSOA is denoted as Pin, and the performances of the received packets are evaluated by a bit-error-rate tester (BERT) and an oscilloscope visualizer, from which the Q factor, bit-error rate (BER), eye-diagram and the waveform of the converted electrical signals at the receiver end are monitored. It should be noted that the decision threshold is optimized for each packet in the receiver, that is to say, a burst mode receiver with adequate input power dynamic range is adopted in the simulation, while the power equalization capability is evaluated by packet-to–packet optical power variation, Q factor and BER.
3.1 All-optical power equalization scheme using a TW-SOA or a common RSOA
First, we compare an all-optical equalization scheme using a common TW-SOA with that one using a common RSOA. In the TW-SOA, the reflectivity of the input facet and output facet, R1 and R2, are both 10−5; while R1 and R2 of the RSOA are 10−5 and 1, respectively. Except for the difference in R2, the two SOAs have identical material and structural parameters, and they are also under the same biased condition. The cavity length is 1200 μm and bias current is 400 mA. Figures 5(a) and 5(b) present the gain as a function of packet input power Pin for the TW-SOA and the common RSOA, respectively. Compared with the TW-SOA, the RSOA has much bigger gain and much steeper gain-Pin curve due to double-path optical gain. It clearly shows that the RSOA can be easily driven into saturation regime even at smaller input power.
Figures 6(a) and 6(b) give the output optical power as a function of packet input power Pin. For the TW-SOA, packets at input power levels of −25 and −15 dBm exhibit output power levels of −4.8 and 4.6 dBm, respectively, and the packet-to-packet optical power variation of 10 dB before the TW-SOA becomes 9.4 dB after the TW-SOA. Obviously, a single TW-SOA is not suitable for all-optical power equalization unless two SOAs are used with the first SOA preamplifying the optical signal to drive the second SOA into deep saturation regime. In contrast, for the RSOA, the packets at input power levels of −25 and −15 dBm have output power levels of 10.5 and 8.8 dBm, showing the 10-dB packet power range before the RSOA is reduced to only 1.7 dB after the RSOA, and this is much smaller than its TW-SOA counterpart. The significant reduction of power fluctuation is owing to that the RSOA has worked in deeper saturation for the same input power range.
Figure 7 present the Q factor and bit-error-rate (BER) of the received packet as a function of Pin; and the insets depict the corresponding eye diagrams of the converted electrical signal when Pin is −25 and −15 dBm, respectively. The optical signal from the TW-SOA has higher Q factor, lower BER and more open eye than that from the RSOA. In addition, for the TW-SOA, the Q factor first increases and then drops with increasing Pin. The reason is as follows: when Pin is low, the TW-SOA works far away from gain saturation regime, the Q factor rises with increasing Pin since the output power is increased. However, when Pin becomes higher, pattern effect begins to appear, then the Q factor decreases. By contrast, for the RSOA, the Q factor monotonously drops with increasing Pin because the RSOA already works in saturation regime. Even at small input power, the Q factor is smaller than 6 dB and BER is higher than 10−9. Since Q factor more than 6 is generally required to ensure BER less than 10−9 in optical communication system, forward error correction (FEC) or other coding technique may have to be adopted if a RSOA-based all-optical packet power equalizer is used.
3.2 All-optical power equalizer using a single two-section RSOA
As stated above, there is a trade-off between small packet-to-packet power variation and high quality signal in an all-optical equalizer using a common RSOA. This problem can be solved by using a two-section RSOA. Let’s consider a two-section RSOA with L1/L2 = 1/9 and total cavity length L of 1200 μm. The length of the insulation channel L3 is 20 μm, and the total bias current I (I = I1 + I2) of the RSOA is 400 mA.
Figures 8 (a1) and 8(b1) demonstrate the received optical power, Q factor of the converted electrical signal at the receiver end as a function of current density ratio J1/J2 when the input optical power of the packet are −25 dBm (marked by solid line) and −15 dBm (marked by dotted line), respectively; and Fig. 8(c1) gives the packet-to–packet optical power variation after the RSOA. For the purpose of comparison, the corresponding results of a two-section TW-SOA are also shown in Figs. 8(a2)-8(c2).
When J1/J2 is 1, bias current densities in the two sections are equal to each other and current is injected almost uniformly along the cavity except in the non-biased insulation channel. Figure 8(a1) shows that the maximal output power of the RSOA are obtained not at J1/J2 of 1 but at J1/J2 bigger than 1. When the input power is −25 dBm, the maximum locates at J1/J2 of 1.35; when the input power is −15 dBm, the maximal point shifts to J1/J2 of 1.47. As stated previously, a two-section RSOA can be equivalent to three cascaded SOAs (i.e., section 1 biased at J1 acts as a preamplifier when the optical signals propagate forward, section 2 biased at J2 is the second SOA and it operates as a saturated SOA, and section 1 biased at J1 acts as the third SOA when the optical signals propagate backward in section 1). And when J1/J2 is bigger than 1, the third SOA has higher current density, which is beneficial to higher optical gain, and similar result has been reported in multi-section TW-SOAs [8,14,15].
It can be seen from Fig. 8(b1) that when J1/J2 is more deviated from 1, Q factor becomes larger, indicating better signal quality is obtained at nonuniform current injection, just as predicted previously; and Q factor over 6 dB is achieved in some J1/J2 regions even at large input power of −15 dBm. Furthermore, the packet-to-packet power fluctuation after the RSOA is also small (< 2dB) in these J1/J2 regions, as indicated in Fig. 8(c1).
By contrast, the two-section TW-SOA gets maximal output power when J1/J2 is smaller than 1 as shown in Fig. 8(a2). Under this bias condition, the section near the output facet has higher current density and offers more carriers to the optical signal. Comparing Fig. 8(b2) with Fig. 8(b1), we find that the two-section TW-SOA has bigger Q factor than the two-section RSOA. Furthermore, when the input power is −25 dBm, the Q factor first increases and then decreases as J1/J2 is increased. This changing trend is different from that when the input power is −15 dBm, and it is also different from that when the two-section RSOA is used. The reason lies in the degrees of gain saturation and pattern effect. When the input power is −25 dBm, the TW-SOA is far away from gain saturation, pattern effect is negligible, and then the Q factor is mainly dependent on the output power. On the other hand, when the RSOA is used or an optical signal of 15-dBm input power enters the TW-SOA, gain saturation becomes deeper, and then the Q factor relies on pattern effect rather than output power.
Figure 8(c2) gives the power fluctuation versus J1/J2 of the TW-SOA. The curve exhibits a peak when J1/J2 is smaller than 1, and the peak power fluctuation and the maximal output power occur nearly at the same J1/J2. Furthermore, power fluctuation is rather large, ranging around 10.3 dB when J1/J2 varies. This is because a single TW-SOA is not so easy to work in saturation regime, as stated above.
Figure 9 compares the waveforms of the received packets at different J1/J2 when the two-section RSOA is used and Pin is −25 dBm. As J1/J2 is increased from 1 to 4, nonuniform current injection and less current density in 2nd section slow down the sharp growth of the signal photons and retard deep saturation in the RSOA. As a consequence, signal quality is enhanced due to the suppression of waveform distortion, the reduction of pattern effect.
Figure 10 presents the BER as a function of the received power at the receiver when the two-section RSOA is biased at J1/J2 of 1.9. The input powers of the two consecutive optical packets are −25 dBm and −15 dBm, respectively. The received power is changed by another variable optical attenuator before the receiver. The dotted line is for the received signal when the RSOA is involved and the solid line is for the back-to-back signal (i.e., without the 20-km fiber and the RSOA). The insets in Fig. 10 are the time-domain waveforms of the optical packet before and after the RSOA. It shows that the packet-to-packet power fluctuation is 10 dB before the RSOA, and it is reduced to 1.9 dB after the RSOA. The power penalty is found to be 4.2 dB, and there is a bit-error floor when the received power is over −18 dBm. Though the two-section RSOA biased at proper J1/J2 suppresses pattern effect to some degree, the dispersion effect and nonlinearity in the fiber and the incomplete elimination of the pattern effect in the RSOA results in larger power penalty.
3.3 Influence of the structural parameters and driven current
The total cavity length L (L = L1 + L2 + L3), the length ratio of the two sections and the reflectivity R2 of the rear facet are among the most important structural parameters that characterize a two-section RSOA. In this subsection, we investigate the performance dependence of the power equalizer on these structural parameters and total driven current.
3.3.1 Influence of total cavity length
Figure 11(a) gives the packet-to-packet power fluctuation as a function of L when two optical packets with Pin of −25 dBm and −15 dBm are input into the RSOA, and Fig. 11(b) gives the Q factor and BER of the received packet when Pin is −15 dBm. In the simulation, L1/L2 = 1/9, J1/J2 = 1.9. It should be noted that the total current density is kept constant at 1.67 × 104A/cm2 when L varies. It can be seen from Fig. 11 that larger cavity length helps reduce power variation, but it simultaneously worsens signal quality. This is because SOA with longer cavity is more likely to experience gain saturation, so there is trade-off between small power variation and high signal quality.
3.3.2 Influence of section length ratio
For a two-section RSOA with fixed length L, when L1/L is changed, both the lengths of the two sections and the location of the insulation channel are varied. As a result, the distribution status of photon density and carrier density are changed, thereby the degree of gain saturation is altered, too. For a two-section RSOA with fixed length L of 1200 μm, we simulate the power equalization performance at different L1/L when driven current is 400 mA and J1/J2 is 1.9.
Figure 12(a) gives the packet-to-packet power variation as a function of L1/L when two optical packets with Pin of −25 dBm and −15 dBm are input into the RSOA; and Fig. 12(b) presents the Q factor and BER of the received packets at Pin of −15 dBm. Generally speaking, when the first section has smaller length, the optical equalizer exhibits smaller power variation. And moderate L1/L is good for higher Q factor and smaller BER, but power fluctuation becomes larger.
3.3.3 Dependence on reflectivity R2
In the previous simulations, the reflectivity of the rear facet, R2, is set at 100%. Here we study the effect of R2 on the performance of the power equalizer. Figure 13(a) shows the packet-to–packet power variation against R2 when two optical packets with Pin of −25 dBm and −15 dBm are input into the RSOA; and Fig. 13(b) presents the Q factor and BER of the received packets at Pin of −15 dBm. In the simulation, L = 1200 μm, I = 400 mA, J1/J2 = 1.9. It can be seen from Fig. 13 that both packet-to-packet variation and Q factor reduce with the increase of R2. This is because more optical signals are reflected back into the RSOA by the rear facet and the RSOA is more easily driven into saturation region. Deeper saturation leads to smaller power variation between packets and also causes more serious pattern effect, waveform distortion and ER degradation. Therefore, one has to trade off between power variation and signal quality when choosing reflectivity R2.
3.3.4 Effect of total driven current
Figure 14(a) presents the dependence of packet-to-packet power fluctuation on total driven current I when two optical packets with Pin of −25 dBm and −15 dBm are input into the RSOA, and Fig. 14(b) gives the Q factor and BER of the received packet when Pin is −15 dBm. In the simulation, L = 1200 μm, L1/L2 = 1/9, J1/J2 = 1.9, R2 = 1. Figure 14 shows that bigger driven current leads to smaller packet-to-packet power fluctuation, and it also improves Q factor and reduces BER. On one hand, the RSOA gets saturated more easily at bigger I; on the other hand, pattern effect is suppressed since more carriers are supplied to compensate for the carrier depletion by the optical signal. As a result, power fluctuation drops and Q factor rises with increasing I.
A simple all-optical power equalization scheme based on a single two-section RSOA is proposed. On one hand, double-path optical gain saturates the RSOA easily, thus avoiding the use of an additional preamplifying SOA; and nonuniform injected current density in the two sections is effective for suppressing pattern effect and improving signal quality. On the other hand, a two-section RSOA biased properly is equivalent to three cascaded SOAs, including a preamplifying SOA, a gain-saturated SOA and another SOA. Therefore, smaller packet-to-packet power fluctuation and better signal quality can be obtained in the all-optical power equalizer based on a two-section RSOA. Compared with the scheme which uses cascaded multiple TW-SOAs, a two-section RSOA-based power equalizer needs only a two-section RSOA together with a passive optical circulator to separate the output signals from the input signals, and the cost is lower and maintenance is simpler.
This work was supported by the National High Technology Research and Development Program of China (No. 2013A014401), Specialized Research Fund for the Doctoral Program of Higher Education (SRFDP) (No. 20120142110064) and Natural Science Foundation of Hubei Province (No. 2012FFB02209).
References and links
1. M. Zirngibl, “An optical power equalizer based on one Er-doped fiber amplifier,” IEEE Photon. Technol. Lett. 4(4), 357–359 (1992). [CrossRef]
2. D. Chiaroni, N. Le Sauze, T. Zami, and J.-Y. Emery, “Semiconductor optical amplifiers: a key technology to control the packet power variation,” in Proc. 27th Eur. Conf. on Opt. Comm. (Amsterdam, 2001), 3, 314–315.
3. R. Sato, T. Ito, Y. Shibata, A. Ohki, and Y. Akatsu, “40-Gb/s burst-mode optical 2R regenerator,” IEEE Photon. Technol. Lett. 17(10), 2194–2196 (2005). [CrossRef]
4. N. Cheng, S.-H. Yen, J. Cho, Z. Xu, T. Yang, Y. Tang, and L. G. Kazovsky, “Long reach passive optical networks with adaptive power equalization using semiconductor optical amplifiers,” in Asia Communications and Photonics Conference and Exhibition, (Shanghai, 2009), FS4.
5. L. Liu, C. Michie, A. E. Kelly, and I. Andonovic, “Packet equalization in PONs using adjustable gain-clamped semiconductor optical amplifiers (AGC-SOA),” in Photonics Global Conference, (Singapore, 2010), Tu.B6.4.
6. S. V. Pato, R. Meleiro, D. Fonseca, P. André, P. Monteiro, and H. Silva, “All-optical burst-mode power equalizer based on cascaded SOAs for 10-Gb/s EPONs,” IEEE Photon. Technol. Lett. 20(24), 2078–2080 (2008). [CrossRef]
7. G. T. Kanellos, N. Pleros, D. Petrantonakis, P. Zakynthinos, H. Avramopoulos, G. Maxwell, and A. Poustie, “40 Gb/s 2R burst mode receiver with a single integrated SOA-MZI switch,” Opt. Express 15(8), 5043–5049 (2007). [CrossRef] [PubMed]
8. C. H. Chen, U. Koren, R. E. Behringer, M. Chien, B. I. Miller, and K. F. Dreyer, “Two section semiconductor optical amplifier as optical power equalizer with high output power,” in Proc. Conf. Optical Amplifier and Their Applications, 1998, paper MC3–1.
9. L. Huang, D. Huang, J. Sun, and D. Liu, “Spectral broadening of ultrashort optical pulse due to birefringence in semiconductor optical amplifiers,” Opt. Commun. 223(4-6), 295–300 (2003). [CrossRef]
10. P. Tian, L. Huang, W. Hong, and D. Huang, “Pattern effect reduction in all-optical wavelength conversion using a two-electrode semiconductor optical amplifier,” Appl. Opt. 49(26), 5005–5012 (2010). [CrossRef] [PubMed]
11. M. L. Nielsen, J. Mørk, R. Suzuki, J. Sakaguchi, and Y. Ueno, “Experimental and theoretical investigation of the impact of ultra-fast carrier dynamics on high-speed SOA-based all-optical switches,” Opt. Express 14(1), 331–347 (2006). [CrossRef] [PubMed]
12. A. E. Willner and W. Shieh, “Optimal spectral and power parameters for all-optical wavelength shifting: single stage, fanout, and cascadability,” J. Lightwave Technol. 13(5), 771–781 (1995). [CrossRef]
13. L. V. T. Nguyen, A. J. Lowery, P. C. R. Gurney, and D. Novak, “A time-domain model for high-speed quantum-well lasers including carrier transport effects,” IEEE J. Sel. Top. Quantum Electron. 1(2), 494–504 (1995). [CrossRef]
14. R. Lennox, K. Carney, R. Maldonado-Basilio, S. Philippe, A. L. Bradley, and P. Landais, “Impact of bias current distribution on the noise figure and power saturation of a multicontact semiconductor optical amplifier,” Opt. Lett. 36(13), 2521–2523 (2011). [CrossRef] [PubMed]
15. J. Y. Emery, B. Lavigne, C. Porcheron, C. Janz, F. Dorgeuille, F. Pommereau, E. Gaborit, I. Guillemot-Neubauer, and M. Renaud, “Two-section semiconductor optical amplifier power equalizer with 8dBm output saturation power for 10Gbit/s applications,” in Conference on Optical Amplifiers and their Applications, Technical Digest Series (Optical Society of America, 1999), 30: 179–182.