We observe a near-ideal high speed amplitude impulse response in an SOA-EAM-SOA configuration under optimum conditions. Full amplitude recovery times as low as 10ps with modulation depths of 70% were observed in pump-probe measurements. System behavior could be controlled by the choice of signal wavelength, SOA current biases and EAM reverse bias voltages. Experimental data and impulse response modelling indicated that the slow tail in the gain response of first SOA was negated by a combination of cross-absorption modulation between pump and modulated CW probe, and self-gain modulation of the modulated CW probe in both the EAM and second SOA.
©2012 Optical Society of America
It is anticipated that semiconductor optical amplifiers (SOAs) may perform optical signal processing (OSP) functions such as demultiplexing, format conversion and wavelength conversion in future high-speed optical time-domain multiplexed (OTDM) communication networks . However, the carrier recovery time of an SOA limits the rate at which it can process high-speed signals. In general, the impulse response has an ultrafast component dependent on intraband effects such as spectral hole burning, carrier heating and carrier cooling followed by a slower, interband, band-filling component [1–3].
Many experimental and theoretical studies of the physics of gain and phase dynamics in SOAs have been conducted [4–8]. Despite ultrafast processes in the carrier recovery, full gain recovery in SOAs is ultimately limited by the slow band-filling process. Various methods have been developed to improve the switching capability of SOAs including narrow bandpass filtering, the use of a holding beam, p-doping the SOA active region, concatenating the SOA with an electro-absorption modulator (EAM), and employing two SOAs in a “push-pull” interferometer or the Turbo-Switch configuration [9–26].
An SOA and an EAM have opposing modulation characteristics so the slow tail in the SOA gain recovery can, in principle, be compensated by a concatenated SOA-EAM when operated in either co-propagating or counter-propagating mode [20,21]. Wavelength conversion, waveform distortion compensation, 2R regeneration and spectral shaping based on the concatenated SOA-EAM system have been demonstrated [21–25].
In this paper, we present a modification of the original Turbo-Switch design. The Turbo-Switch consists of two SOAs in series separated by a wide bandpass filter . When a continuous wave (CW) probe and pump pulses are injected into the first SOA, cross-gain modulation (XGM) between the pump and probe occurs. The filter blocks the pump from entering the second SOA, but the modulated probe enters the second SOA and undergoes self-gain modulation (SGM). This leads to elimination of the slow tail in the gain recovery of the first SOA and creates an overshoot in the amplitude recovery . Here, we replace the tunable bandpass filter with an EAM to prevent the transmission of pump pulses to the second SOA, allowing for the possibility of integration of the concatenated SOA-EAM-SOA (CSES) on a single chip. Devices consisting of SOAs integrated with EAMs have already been fabricated and employed in optical access networks [27,28] and 2R regeneration [22,24].
We investigated cross-amplitude modulation in the CSES configuration operated in co-propagating mode. Pump-probe measurements using a 3ps pump and a CW probe on the CSES revealed amplitude recovery that was highly dependent on the CW probe wavelength, on the current biases of the SOAs and on the EAM reverse bias voltage. When these parameters were optimized, very fast amplitude recovery (within 10ps) without a significant overshoot could be achieved. This remarkably fast amplitude response could potentially be exploited for use in high-speed (≥100Gb.s−1) optical signal processing operations.
The layout of this paper is as follows: Section 2 provides details of the experimental techniques used and the experimental data from these experiments is described in Section 3. Section 4 contains a description of the phenomenological impulse response model which accounts for the behavior of the CSES. The principal findings of this work are summarized in Section 5.
2. Experimental details
In order to measure the amplitude response of individual components of the CSES as well as the entire CSES, the pump-probe experiment shown schematically in Fig. 1 was used.
When measurements on an individual component (e.g. a single SOA or EAM) took place, only that component remained in the test-bed while all other components were removed from the section between the 50:50 coupler and the first bandpass filter (BPF) in Fig. 1. Pump pulses and a CW probe beam were combined using a 50:50 coupler before entering the device(s) under test. 10% of the power from the pump pulse train was sent directly to the oscilloscope to enable it to trigger off a 10.65GHz external clock. The signal from the device(s) under test was amplified using an erbium-doped fiber amplifier (EDFA), whose filtered output was sent into a 90:10 coupler before entering the oscilloscope.
The pump was a 1535nm 3ps pulse clock stream, with a 3dB bandwidth of 1.68nm, from an actively mode-locked tunable laser, whose repetition rate was reduced from 10.65GHz to 665MHz using a LiNbO3 optical modulator. The CW probe wavelength was varied within the range 1545≤ λprobe≤1565nm; this ensured a wavelength separation between pump and probe of at least 10nm (the longest wavelength that could be detected by the OSO was 1565nm). A 4nm band-pass filter after the device(s) under test removed both amplified spontaneous emission (ASE) and pump signals. The output probe signal passed through an EDFA to a 500GHz optical sampling oscilloscope. For time-resolved measurements of the CSES, the CW probe input average power was fixed at −4.3dBm and the pump energy per pulse was maintained at 90fJ to ensure a large modulation depth (≥50%).
The two SOAs employed were commercially available InGaAsP/InP buried heterostructure multiple quantum well (MQW) nonlinear devices with nearly identical properties. For each SOA, the confinement factor was 0.4, the effective length was 1mm and input and output coupling losses varied from 2.0 to 3.9dB per facet. With a current bias of 200mA on each SOA, the 3dB gain bandwidth was 80nm, the small signal gain peak was 1510nm, the saturation input power was −19dBm and the fiber-to-fiber (FtF) small-signal gain was 29dB. The FtF saturated gain under typical experimental conditions when there was a saturating CW beam incident on each SOA were 19dB and 14dB for SOA 1 and SOA 2 respectively. The SOA current biases were varied from 100mA to 400mA.
The EAM used in the CSES configuration was a commercially-available InGaAsP/InP buried heterostructure MQW EAM with an effective length of 0.3mm, an input coupling loss of 3.1dB, an output coupling loss of 2.9dB and the band-edge was approximately 1500nm. The reverse bias voltage applied to the EAM was varied from 0 to 5V.
Figure 2(a) shows the small signal chip gain spectrum for SOA 1 with a current bias of 200mA. The series of spikes in the spectrum near the gain peak were due to gain ripple. Figure 2(b) shows the chip absorption spectra for the EAM as the applied reverse voltage bias was varied from 0 to 5V, showing the customary red-shift due to the quantum-confined Stark effect (QCSE) .
3. Experimental results
Typical normalized gain evolutions for SOA 1 with the current bias varied from 100 to 400mA with the CW probe set to 1565nm are shown in Fig. 3 . The amplitude evolutions were normalized relative to the steady-state amplitude level when no pump pulse was incident on the device(s) under test. The average CW power input and the energy per pump pulse at the input to SOA 1 were −4.3dBm and 90fJ respectively. The subsidiary peaks that appeared in the amplitude evolutions at intervals of 94ps were due to 10.65GHz components which were not fully suppressed by the LiNbO3 modulator that reduced the pump repetition rate to 665MHz. Both ultrafast and slow intraband components were visible in the evolutions and the full recovery time was ~150ps (the slow tail had a fitted 1/e recovery time of 38ps) at a current bias of 400mA.
Typical normalized transmission responses of the EAM for a 1565nm CW probe under various reverse voltage bias conditions are shown in Fig. 4 , where the pump pulse energy was 0.6pJ and the CW probe power was 10.8dBm. These input powers were selected because when the EAM was part of the CSES configuration, the energy per pump pulse varied from 0.30 to 0.75pJ and the average probe power varied from 8 to 11.8dBm at the output port of SOA 1 as the SOA drive current varied from 100 to 400mA.
The probe amplitude response of SOA 1 in series with the EAM is shown in Fig. 5 with SOA current biases of 100mA and 400mA.
As can be seen in Fig. 5, at EAM reverse bias voltages of ≤1.0V, the start of the slow tail in the gain evolution of SOA 1 was partially compensated. However, as the EAM bias increased to ≥2.3V, there was a fast initial recovery followed by a sharp overshoot, a subsequent dip below the steady-state level and a long recovery tail in the amplitude evolution.
Figure 6 shows the amplitude evolution of the entire CSES configuration for both low (100mA) and high (400mA) currents on both SOAs.
The graphs in Fig. 6 show that for 100mA current biases for both SOAs and no EAM voltage bias, there was only a small degree of compensation for the slow tail in the total response of the CSES. However, for 400mA SOA current biases for both SOAs and an EAM voltage bias of 2.3V, the amplitude response consisted predominantly of an ultrafast component as the slow band-filling tail was almost completely eliminated.
In fact, it was possible to achieve near-total compensation for the slow band-filling tail as long as the SOA 1 current was ≥300mA, the SOA 2 current was ≥200mA and the EAM bias was within the range 2.1≤VEAM≤2.5V. Two optimized amplitude evolutions of the CSES with are displayed in Fig. 7 , both of which show full amplitude recovery times of ~10ps and negligible overshoot.
In order to illustrate the effect of the EAM and SOA 2 on the output from SOA 1, the modulated CW signal as it progressed through the CSES system is displayed in Fig. 8 for the case when the bias currents for SOA 1 and SOA 2 are 300mA and 200mA respectively.
In Fig. 8(a), where no voltage was applied to the EAM, the CSES behaved similarly to two SOAs in series. However, in Fig. 8(b), where a non-zero voltage bias was applied to the EAM, both the EAM and SOA 2 changed the amplitude evolution of the probe.
Referring back to Fig. 2(b), we note that the pump absorption by the EAM was weak until the bias voltage exceeded ~1.0V. Above this voltage, the pump was strongly attenuated and the probe was moderately attenuated by the EAM. The strong absorption of the pump had two effects: (i) it enabled cross-absorption modulation (XAM) between pump and probe in the EAM and (ii) it prevented transmission of most of the pump power to SOA 2, thereby allowing SGM of the probe to dominate over XGM between the pump and probe in SOA 2. The combination of XAM between the pump and probe in the EAM and SGM of the probe in SOA 2 led to the remarkably fast amplitude recovery of the CSES under optimum conditions. However, as the EAM bias increased beyond 2.5V, the absorption of the probe power by the EAM increased, causing self-amplitude modulation of the probe to take place in the EAM, thus creating a dip following the overshoot in the amplitude response.
In Fig. 9 , the 10% to 90% (10/90) recovery times are plotted as a function of EAM voltage bias for various SOA currents. The lines between points are present to act as a guide to the eye.
It can be seen from Fig. 9(a) that the CSES recovery time decreased as the SOA 2 current bias increased and as the EAM bias increased. A similar trend can be observed in Fig. 9(b) but the recovery times are all shorter relative to those in Fig. 9(a) due to the larger SOA 1 current bias. The 10/90 recovery times for the CSES when the EAM bias increased beyond 2.3V for Fig. 9(a) and 2.5V for Fig. 9(b) are not shown in the graphs. When the EAM bias was greater than these values, although the initial amplitude response was very fast (<10ps), the prolonged dip in the response caused the amplitude to return below the threshold of 90% full recovery.
Figure 10 shows CSES amplitude evolutions for several probe wavelengths and various values of attenuation on Att. 1. The CSES amplitude evolutions with several values of attenuation on Att. 2 were similar to the evolutions displayed in Fig. 10(b).
Figure 10(a) shows that the CSES amplitude recovery rate improved as the CW wavelength increased. At shorter CW wavelengths, the EAM absorbed more probe power so there was reduced SGM of the probe in SOA 2. At longer CW wavelengths, the EAM absorbed less probe power, leading to increased SGM in SOA 2; therefore, there was greater compensation for the slow tail in the input to SOA 2. Figure 10(b) demonstrates that when the power of the pump and CW probe signals were attenuated, there was a decrease in both XGM between pump and probe in the EAM, and in SGM of the probe in SOA 2, leading to a reduction in the CSES amplitude recovery rate.
4. Theory and modelling
A simple phenomenological model was developed to explain the observed behavior of the CSES configuration. Firstly, the gain coefficient impulse response, gcoeff,x(t), of each SOA was modeled using Eq. (1), where the subscript x refers to either SOA 1 or SOA 2.
For all of the equations in this paper, t refers to time, the subscripts “bf” and “ch” refer to the carrier recovery processes of band-filling and carrier-heating respectively, “τ” refers to the time constants of these recovery processes and “a” determines the contribution of band-filling and carrier heating to the total impulse response. The fitted time constants for band-filling and carrier heating were 38≤τbf≤65ps and 1ps respectively. For each SOA, gcoeff,x(t) was convolved with a modeled sech2 pump pulse, ypump(t) (Eq. (2)).
The convolved gain, gconv,x(t), was converted into a modelled power gain response (Eq. (3)).
Gconv,x(t) was multiplied by a normalized CW input (i.e. unity) to calculate the modeled gain evolution at the probe wavelength. The parameters of ypump(t) and gcoeff,x(t) were chosen to ensure a good fit between the modelled and the measured normalized gain response for each SOA. In Fig. 11 , the experimental gain evolution of SOA 1 is compared with the modeled gain evolution.
Having fitted the power gain response of each individual SOA, the output from SOA 1 + EAM in series was modeled. When an optical pulse is incident on an EAM at a wavelength within the absorption spectrum, electron-hole pairs are created within the active region. Under the influence of the electric field due to the built-in potential and the application of a reverse bias voltage, these carriers escape from the wells and drift towards the n-type and p-type heterojunctions. The carriers accumulate at the heterojunctions until they are swept out. Carrier escape from both the wells and heterojunctions is via thermionic emission and tunnelling [30,31]. The accumulated carriers create a space-charge field which reduces the effective electric field within the active region, causing the absorption spectrum to become blue-shifted .
The absorption coefficient of an EAM is a function of voltage, carrier density, carrier temperature and photon density . However, in order to model the behavior of the EAM in response to the injection of a pump pulse train and a modulated CW probe without using detailed rate equations, it was assumed that the main contribution to the absorption dynamics was from field-induced screening [21,32,33]. The field screening, which directly reduces the effective reverse bias voltage, is proportional to the number of photo-generated carriers. The EAM voltage change impulse response, ΔVEAM(t), was modeled as:
In Eq. (4), av determines the maximum change in effective reverse bias voltage from the steady-state value, τV,rise is the rise time constant, and τV,rec is the carrier sweep-out time constant. The rise time reflects the finite time taken for carriers to escape from the wells and travel to the contacts, which may be affected by the difference in carrier mobility between electrons and holes.
The pump pulse, ypump(t), and the probe output from SOA 1, Gconv,SOA1(t), were each convolved with ΔVEAM(t).The convolutions of the modeled pump pulse and the modulated CW probe with ΔVEAM(t) were summed using a weighting factor which accounted for the smaller probe absorption cf. pump absorption in the EAM. This weighted sum was added to the steady-state reverse bias to calculate the temporal evolution of the effective reverse bias, VEAM(t). In turn, VEAM(t) was used to calculate the EAM transmission at the pump and probe wavelengths. This was done by deriving a static voltage-dependent transmission function from experimental data (see Fig. 12 ).
The calculated transmission was multiplied by the probe input to the EAM, Gconv,SOA1(t), to produce the modeled amplitude evolution of the output from SOA 1 and the EAM in series, XSOA1,EAM(t). Figure 13 shows the modeled EAM input, transmission and output, together with the experimentally measured output from SOA 1 + EAM at the probe wavelength for an SOA bias current of 400mA at two different EAM bias voltages.
Once a good fit between the modeled and measured output from SOA 1 and the EAM in series was obtained, ypump(t) and XSOA1,EAM(t),were both convolved with gcoeff,SOA2(t). These convolutions were added together using weighing factors to take into account the fact that the probe input power was greater than the pump input power and that the pump and CW probe did not experience the same gain in SOA 2. The weighted sum of these convolutions were used to calculate the power gain and this was multiplied by the probe input to SOA 2 to produce the modeled output from the CSES. In Fig. 14 , the modeled SOA 2 input, transmission and output, in addition to the experimentally measured output from the entire CSES configuration at the probe wavelength are displayed.
In Fig. 14(a), it can be seen that when the EAM bias was low (≤1.0V), the transmission of SOA 2 contained contributions from both XGM between the pump and probe, and SGM of the probe within SOA 2. In the modeled transmission of SOA 2, XGM manifests itself as a drop in transmission near t = 0 while SGM manifests itself as a rise in transmission. In Fig. 14(b), where the EAM bias was higher (>1.0V), the pump was absorbed by the EAM and there was negligible XGM between pump and probe in SOA 2. For that reason, the transmission of SOA 2 consisted almost entirely of a component due to SGM of the probe. The good agreement between the modeled and experimentally measured amplitude evolutions demonstrates that the impulse response model accounts for the main features of the CSES amplitude response.
The amplitude dynamics of a concatenated SOA-EAM-SOA configuration have been investigated using time-resolved pump-probe spectroscopy. The CSES behavior could be controlled by choice of signal wavelength, SOA current bias and EAM voltage bias. The slow tail in the gain recovery of the output from SOA 1 could be cancelled out partially or fully when suitable operation parameters were chosen.
Under certain operating conditions, full amplitude recovery within 10ps (corresponding to a 1/e recovery time of 2.9ps), consisting predominantly of an ultrafast component without an overshoot was attained. In general, fastest amplitude recovery rates were achieved when the CW probe was at the longest wavelength that could be detected by the optical sampling oscilloscope (1565nm), the current bias for the SOAs was moderate (≥300mA for SOA 1, ≥200mA for SOA 2) and the EAM bias was within the range 2.1≤VEAM≤2.5V. Good agreement could be obtained between the experimental amplitude response and computationally modeled amplitude response using a phenomenological model, indicating that the observed behavior of the CSES resulted from a combination of XAM between pump and probe, and SGM of the probe in both the EAM and SOA 2.
The CSES has several advantages over the conventional Turbo-Switch. The recovery rate of the CSES is faster than that of the Turbo-Switch when the amplitude impulse responses for both configurations are compared (at 70% modulation depth). When the Turbo-Switch was operated under optimum conditions (1550nm pump, 1536nm CW, drive currents of 400mA for both SOAs, energy per pump pulse of 125fJ, average CW power of −0.2dBm), it had a full amplitude recovery time of 30ps . This is longer than the full amplitude recovery time for the CSES of 10ps under optimum operating conditions (mentioned in the previous paragraph, with energy per pump pulse of 90fJ, average CW power of −4.3dBm). Furthermore, the CSES can be monolithically integrated onto a single chip, unlike the Turbo-Switch. One disadvantage of the CSES is sensitivity to the choice of operation parameters.
Although a near-perfect amplitude response could only be achieved within a narrow range of operation parameters for the CSES, the SOA drive currents and the EAM reverse bias can be accurately controlled using stable current and voltage sources. In addition, the EAM employed in experiments described in this paper was not ideal for CW probe wavelengths throughout the C-band (1530 – 1565nm). If the EAM band edge was red-shifted cf. the EAM employed in this paper, the 1535nm pump would be strongly absorbed while the CW probe (1545≤λprobe≤1565nm) would not be strongly absorbed by the EAM at low reverse bias (≤1.0V). This would allow XAM between pump and probe, with negligible self-amplitude modulation of the probe to take place in the EAM, which may lead to a near-perfect amplitude response for a wide range of CW probe wavelengths at low reverse bias.
Based on the experimental work outlined in this paper, an integrated version of the CSES operated under optimum operating conditions could be employed as a compact switching device in high-speed (≥100Gb.s−1) optical networks, as its amplitude recovery time is limited only by ultrafast recovery processes in the SOAs, pump pulse width and modulation depth.
This work was supported by Science Foundation Ireland under grant 06/IN/I969.
References and links:
1. D. Cotter, R. J. Manning, K. J. Blow, A. D. Ellis, A. E. Kelly, D. Nesset, I. D. Phillips, A. J. Poustie, and D. C. Rogers, “Nonlinear optics for high-speed digital information processing,” Science 286(5444), 1523–1528 (1999). [CrossRef] [PubMed]
2. W. Mathlouthi, F. Vacondio, P. Lemieux, and L. A. Rusch, “SOA gain recovery wavelength dependence: simulation and measurement using a single-color pump-probe technique,” Opt. Express 16(25), 20656–20665 (2008). [CrossRef] [PubMed]
3. L. Occhi, Y. Ito, H. Kawaguchi, L. Schares, J. Eckner, and G. Guekos, “Intraband gain dynamics in bulk semiconductor optical amplifiers: measurements and simulations,” IEEE J. Quantum Electron. 38(1), 54–60 (2002). [CrossRef]
4. R. Giller, R. J. Manning, and D. Cotter, “Gain and phase recovery of optically excited semiconductor optical amplifiers,” IEEE Photon. Technol. Lett. 18(9), 1061–1063 (2006). [CrossRef]
5. F. Girardin, G. Guekos, and A. Houbavlis, “Gain recovery of bulk semiconductor optical amplifiers,” IEEE Photon. Technol. Lett. 10(6), 784–786 (1998). [CrossRef]
6. R. Gutierrez-Castrejon, L. Schares, L. Occhi, and G. Guekos, “Modeling and measurement of longitudinal gain dynamics in saturated semiconductor optical amplifiers of different length,” IEEE J. Quantum Electron. 36(12), 1476–1484 (2000). [CrossRef]
7. L. Schares, C. Schubert, C. Schmidt, H. G. Weber, L. Occhi, and G. Guekos, “Phase dynamics of semiconductor optical amplifiers at 10-40 GHz,” IEEE J. Quantum Electron. 39(11), 1394–1408 (2003). [CrossRef]
8. A. Uskov, J. Mork, and J. Mark, “Theory of short-pulse gain saturation in semiconductor laser amplifiers,” IEEE Photon. Technol. Lett. 4(5), 443–446 (1992). [CrossRef]
9. M. Asghari, I. H. White, and R. V. Penty, “Wavelength conversion using semiconductor optical amplifiers,” J. Lightwave Technol. 15(7), 1181–1190 (1997). [CrossRef]
10. J. Leuthold, R. Ryf, D. N. Maywar, S. Cabot, J. Jaques, and S. S. Patel, “Nonblocking all-optical cross connect based on regenerative all-optical wavelength converter in a transparent demonstration over 42 nodes and 16800 km,” J. Lightwave Technol. 21(11), 2863–2870 (2003). [CrossRef]
11. 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, “Error-free 320-Gb/s all-optical wavelength conversion using a single semiconductor optical amplifier,” J. Lightwave Technol. 25(1), 103–108 (2007). [CrossRef]
13. R. J. Manning, D. A. O. Davies, S. Cotter, and J. K. Lucek, “Enhanced recovery rates in semiconductor laser amplifiers using optical pumping,” Electron. Lett. 30(10), 787–788 (1994). [CrossRef]
14. R. J. Runser, D. Zhou, C. Coldwell, B. C. Wang, P. Toliver, K. L. Deng, I. Glesk, and P. R. Prucnal, “Interferometric ultrafast SOA-based optical switches: From devices to applications,” Opt. Quantum Electron. 33(7/10), 841–874 (2001). [CrossRef]
15. J. P. Sokoloff, P. R. Prucnal, I. Glesk, and M. Kane, “A terahertz optical asymmetric demultiplexer (TOAD),” IEEE Photon. Technol. Lett. 5(7), 787–790 (1993). [CrossRef]
16. K. E. Stubkjaer, “Semiconductor optical amplifier-based all-optical gates for high-speed optical processing,” IEEE J. Quantum Electron. 6(6), 1428–1435 (2000). [CrossRef]
17. G. Talli and M. J. Adams, “Gain recovery acceleration in semiconductor optical amplifiers employing a holding beam,” Opt. Commun. 245(1-6), 363–370 (2005). [CrossRef]
18. L. Zhang, I. Kang, A. Bhardwaj, N. Sauer, S. Cabot, J. Jaques, and D. T. Neilson, “Reduced recovery time semiconductor optical amplifier using p-type-doped multiple quantum wells,” IEEE Photon. Technol. Lett. 18(22), 2323–2325 (2006). [CrossRef]
19. L. Zhang, I. Kang, A. Bhardwaj, N. Sauer, S. Cabot, J. Jaques, and D. T. Neilson, “Significant reduction of recovery time in semiconductor optical amplifier using p-type modulation doped MQW,” in European Conference on Optical Communications (ECOC,2006), 1–2.
20. Y. Yu, L. Huang, M. Xiong, L. Yan, and P. Tian, “An integrated circuit subsystem of quantum dot semiconductor optical amplifier coupled with electro-absorption modulator and its application in wavelength conversion,” Opt. Commun. 284(7), 1847–1854 (2011). [CrossRef]
21. E. Zhou, F. Öhman, C. Cheng, X. Zhang, W. Hong, J. Mørk, and D. Huang, “Reduction of patterning effects in SOA-based wavelength converters by combining cross-gain and cross-absorption modulation,” Opt. Express 16(26), 21522–21528 (2008). [CrossRef] [PubMed]
22. M. J. R. Heck, A. J. M. Bente, Y. Barbarin, D. Lenstra, and M. K. Smit, “Monolithic semiconductor waveguide device concept for picosecond pulse amplification, isolation, and spectral shaping,” IEEE J. Quantum Electron. 43(10), 910–922 (2007). [CrossRef]
23. K. Inoue, “Technique to compensate waveform distortion in a gain-saturated semiconductor optical amplifier using a semiconductor saturable absorber,” Electron. Lett. 34(4), 376–378 (1998). [CrossRef]
24. F. Ohman, R. Kjær, L. J. Christiansen, K. Yvind, and J. Mork, “Steep and adjustable transfer functions of monolithic SOA-EA 2R regenerators,” IEEE Photon. Technol. Lett. 18(9), 1067–1069 (2006). [CrossRef]
25. T. Vivero, N. Calabretta, and I. T. Monroy, G. CarvalhoKassar, F. Ohman, K. Yvind, A. Gonzalez-Marcos, and J. Mork, “10 Gb/s-NRZ optical 2R-regeneration in two-section SOA-EA chip,” in IEEE conference on Lasers and Electro-Optics Society (LEOS,2007), 806–807.
26. R. P. Giller, R. J. Manning, and D. Cotter, “Recovery dynamics of the turbo-switch,” in Optical Amplifiers and Their Applications, Technical Digest (Optical Society of America, 2006), paper OTuC2.
27. E. K. MacHale, G. Talli, P. D. Townsend, A. Borghesani, I. Lealman, D. G. Moodie, and D. W. Smith, “Extended-reach PON employing 10Gb/s integrated reflective EAM-SOA,” in European Conference on Optical Communications (ECOC,2008), 1–3.
28. G. Talli and P. D. Townsend, “Hybrid DWDM-TDM long-reach PON for next-generation optical access,” J. Lightwave Technol. 24(7), 2827–2834 (2006). [CrossRef]
29. G. L. Li and P. K. L. Yu, “Optical intensity modulators for digital and analog applications,” J. Lightwave Technol. 21(9), 2010–2030 (2003). [CrossRef]
30. S. Hojfeldt and J. Mork, “Modeling of carrier dynamics in quantum-well electroabsorption modulators,” IEEE J. Quantum Electron. 8(6), 1265–1276 (2002). [CrossRef]
31. A. V. Uskov, J. R. Karin, R. Nagarajan, and J. E. Bowers, “Dynamics of carrier heating and sweepout in waveguide saturable absorbers,” IEEE J. Quantum Electron. 1(2), 552–561 (1995). [CrossRef]
32. J. R. Karin, R. J. Helkey, D. J. Derickson, R. Nagarajan, D. S. Allin, J. E. Bowers, and R. L. Thornton, “Ultrafast dynamics in field‐enhanced saturable absorbers,” Appl. Phys. Lett. 64(6), 676–678 (1994). [CrossRef]