Open Access
Add to BibSonomyAbstract
We propose a novel approach for simultaneously controlling the chirp and increasing the output power of an EADFB laser by monolithically integrating a short-cavity SOA. We achieved a 40-Gbit/s 5-km SMF transmission at a wavelength of 1.55 μm by using an EADFB SOA with a lower power consumption than a stand-alone EADFB laser.
© 2015 Optical Society of America
1. Introduction
The continuing increase in data traffic caused by the spread of, for example, cloud computing and smart phones requires an optical network system with a higher capacity. In data centers, there is a real need for optical network systems to increase their transmission capacity with reduced power consumption. This means that the signaling rate of the optical light source used for this application must be increased and its power consumption simultaneously reduced.
An electroabsorption (EA) modulator integrated with a distributed feedback (EADFB) laser is widely used for many applications because of its large extinction ratio, and small chirp parameter in both the 1.3- and 1.55-μm wavelength windows. High-speed operation up to 50 Gbit/s targeted at 400GbE applications has been demonstrated in the 1.3-μm wavelength region [1]. In the 1.55-μm wavelength region, it is used for 10-Gbit/s 40- and 80-km SMF transmission and 40-Gbit/s 2-km SMF transmission [2,3]. As regards these applications, there have been some recent reports on an EADFB laser where the injection current of the DFB laser section (ILD) is less than 80 mA to reduce its power consumption [4.5]. To further reduce power consumption and cope with the continuing increase in data traffic, the efficiency of the optical output power to ILD must be increased and its chirp reduced simultaneously with low power consumption, because this can extend the transmission distance. In addition, a feature of a high output power is that it is suitable for operation not only with a conventional non-return-zero modulation format but also with new modulation formats such as pulse amplitude modulation (PAM), which make it possible to utilize the frequency bandwidth of the light source efficiently. However, with an EADFB laser it is difficult to increase the output power and reduce chirp simultaneously, i.e. in principle there is a limitation that prevents us from designing the refractive index change and the absorption coefficient independently, and this is because of the Kramers-Kronig (K-K) relation of the EA modulator. This relationship inevitably requires a large (small) optical loss when the chirp parameter is designed to be small (large). And this cannot be overcome solely by optimizing the multi-quantum-well (MQW) structure.
With a view to overcoming this limitation of the EA modulator, there have been some reports that use a dual modulation technique [4], or chirp management with a semiconductor optical amplifier (SOA) [5–8]. The dual modulation technique can extend the transmission distance by modulating the DFB section as well as the EA modulator, but this requires the difficult procedure of RF signal tuning and cannot reduce the power consumption of the EADFB laser. On the other hand, a chirp control technique using an SOA can be realized simply by injecting DC current into the SOA. This is because, when the optical input power to SOA varies, the carrier density into the SOA varies inversely leading to a negative chirp. Figure 1 summarizes reports about an SOA designed to compensate for the loss and chirp of an EA modulator. Although some reports did not specify the SOA length, most used a long cavity SOA with the large injection current for the SOA (ISOA) of over 50 mA. This inevitably results in greater power consumption than that of a stand-alone EADFB laser. Moreover, the reported operating speed remained at 23 Gbit/s. There have been some theoretical studies of high-speed SOA dynamics [9]. If an SOA is used under operating conditions consisting of a high carrier density and a high optical input power, the effective lifetime of the carrier can be reduced and the SOA can respond at over 40 Gbit/s. But there have been no reports of the simultaneous compensation of both chirp and output power using an SOA under operating conditions where the power consumption is lower than that of a stand-alone EADFB laser.
In this paper, to overcome the limitation of the EADFB laser described above, we devise a novel approach that uses two positive effects efficiently provided by an SOA [10]. One is output power compensation, which is achieved by balancing ILD and ISOA and optimizing the total power consumption of the EADFB SOA. The other is chirp compensation, which is achieved by using an SOA to invert the carrier density compared with that of an EA modulator. When ILD decreases, the photocurrent of the EA section (IEA) decreases, and this also results in a reduction in the power consumption of EA section. Taking this into account, we carefully design the SOA length so that the output power is increased and the chirp reduced simultaneously with a lower total power consumption than a stand-alone EADFB laser. We also evaluate its characteristics at 45°C to reduce the power consumption of the thermo-electric cooler. We investigate its dynamics at 40 Gbit/s, and realize a 5-km single mode fiber (SMF) transmission that is conventionally 2-km for EADFB laser.
2. Device design and fabrication
Figure 2 shows a schematic diagram and the driving concept of an EADFB laser integrated with an SOA. To increase the optical output power and reduce the chirp parameter without increasing the power consumption (P) compared with a stand-alone EADFB laser, we focus on the injection current versus optical output power efficiency, namely the slope efficiency, of the LD and SOA section. To achieve a high optical output power with a small injection current, it is important to realize a balance between ILD and ISOA. As described above, in practical use, ILD ranged from 60 to 80 mA [11,12], and therefore we looked for the optimum driving condition for ISOA with an EADFB SOA when ILD was 60 mA. We also set the ILD of the stand-alone EADFB laser at 80 mA, and investigated the length of SOA that was beneficial for the EADFB SOA.
With this motivation as a basis, we integrated an EADFB laser with an SOA (EADFB SOA). This allowed the SOA to operate in a high optical input condition, and resulted in an SOA with a high carrier density. The high carrier density of the SOA will boost the SOA speed, because the effective carrier lifetime will be short. We also employed InGaAlAs material with MQWs for the LD, EA, and SOA section, because this material has the features of high temperature tolerance, and high-speed operation. To avoid the need for a complex fabrication process, we used the same MQW structure that we used for the LD and SOA, and no additional epitaxial regrowth process is needed compared with the stand-alone EADFB laser. InGaAsP passive waveguide is employed between LD and EA section, and EA and SOA section by using butt-joint process. A ridge waveguide structure buried in benzocyclobutene (BCB) was employed to reduce the parasitic capacitance of the EAM section [3]. The front and rear facets were coated with antireflection (AR) and high-reflection (HR) film. The differential resistance of LD section was about 12 Ω.
First, we estimate the SOA gain in relation to SOA length and ISOA parameters by simulating the dynamic behavior of carriers and photons with the help of a rate equation. The rate equation is expressed as,
where N is the carrier density, S is the photon density, I is the injection current of the SOA, e is the electron charge, V is the volume of the SOA, τs is the carrier lifetime, τp is the photon lifetime, vg is the group velocity, Ag is the differential gain, N0 is the carrier density at the transparency point, Γ is the optical confinement factor, and Pin is the optical input. The gain is expressed as,where L is the SOA length. We calculated the gain using the Runge-Kutta method. The Γ value calculated with the finite element method was 0.052. The general values of 1 × 10−9 [s], 1.5 × 10−12 [s], and 9 × 109 [cm/s] were used for τs, τp, and vg.Figure 3 shows the SOA length and ISOA dependence of the calculated gain versus the optical input power. An Ag of 3.0 × 10−16 [cm2] and an N0 of 5.0 × 1018 [cm−3] were used for the calculation. The tendency of the calculation is that when ISOA is small, a shorter SOA can provide a higher gain. As seen in the figure, the gain of a 50-μm-long SOA with an ISOA of 10 mA will exceed 1.1 dB when input power to the SOA is 10 dBm and this is a standard SOA input power when the bias voltage of the EA modulator (Vdc) is in an open circuit condition. Although a longer SOA with an ISOA of 25 mA can provide a larger gain, it will result in increased power consumption. For this reason, the SOA length needs to be minimized as the required gain is obtained. To investigate the validity of this approach, we measured the light-current (L-I) characteristics of a stand-alone EADFB laser and an EADFB SOA with an open Vdc.
3. Device characteristics
Figure 4 shows the measured L-I characteristics of a stand-alone EADFB laser, and an EADFB SOA with SOA lengths of 50, 100, and 150 μm. The ISOA values were set as the current densities of the SOAs became the same, i.e. 10 and 25 mA for a 50-μm-long SOA, and 30 and 75 mA for a 150-μm-long SOA. As shown in the figure, when the SOA is 50 μm, the optical output power at ILD and ISOA values of 57 and 10 mA for an EADFB SOA is same as that of an EADFB laser at an ILD of 80 mA. On the other hands, when the SOA length is 150 μm, the optical output power at ILD and ISOA values of 40 and 30 mA for an EADFB SOA is same as that of an EADFB laser at an ILD of 80 mA. To compare the benefit of SOAs with three different lengths, we measured the gain of three SOAs when ILD was fixed at 60 mA.
Figure 5 shows the measured the ISOA gain dependence for 50-, 100-, and 150-μm-long SOAs with an ILD of 60 mA at 45°C. As indicated in Fig. 4, when ILD was reduced from 80 to 60 mA, the optical output power decreased by 1.3 dB, and the red line in Fig. 5 shows a gain of 1.3 dB. This reveals that the merit of the SOA can be obtained when the gain exceeds 1.3 dB. When we use a 50-μm SOA, an ISOA of over 7 mA will provide a gain of over 1.3 dB.
Next, to determine the maximum ISOA, we measured the power consumption (P) of the SOA. Figure 6 shows the SOA length dependence of the SOA power consumption versus ISOA at 45°C. The bias voltages of 50-, 100-, and 150-μm-long SOAs are about 1.3, 1.2, and 1.2 V when ISOA are 10, 20, and 30 mA, respectively, and this corresponds to SOA resistance of 130, 61, and 41 Ω. When the ILD value was reduced from 80 to 60 mA, the power consumption levels of the LD and EA sections were reduced by about 36 and 16 mW, respectively, because the photo current in the EA modulator decreased with the reduced optical input into the EA modulator. The red line shown in Fig. 6 indicates 52 mW, which is the sum of the reduced P of the LD and EAM section. When the SOA length is 50 μm, an ISOA of below 25 mA will give a smaller P than a stand-alone EADFB laser. From Figs. 5 and 6, the tolerance of the ISOA of the 50-μm-long SOA was sufficiently large, and the value was 18 mA. We can obtain the benefit of the SOA with a tolerance of 18 mA, and this provides us with a very simple operation.
Next, we examined whether or not the amplified spontaneous emission (ASE) of the SOAs pose a problem with respect to the lasing spectra. Figure 7 shows the lasing spectra of EADFB SOAs with three different lengths at 45°C. ILD was set at 60 mA. Although the lasing spectra of the EADFB SOA with an SOA length of 150 μm have a noise level with a non-uniform shape, a side mode suppression ratio (SMSR) of over 50 dB was obtained in each case. These results indicate that the ASE from a 50-μm-long SOA does not have a large impact on the lasing spectra.
To investigate the influence of noise from SOA, we measured the relative intensity noise (RIN) spectra. Figure 8 shows the measured RIN spectra of an EADFB laser and EADFB SOAs whose SOA lengths are 50, 100, and 150 μm. ILD was set at 60 mA. The EAM is open circuit. As shown in these figures, the large difference between the RIN spectra of EADFB laser and that of EADFB SOA is not observed. We consider that the noise at very low frequency is caused by our experimental setup. The low RIN values of less than −140 dB/Hz are measured for each device.
We also investigated the influence of the ASE from an SOA on the static extinction characteristics. Figure 9 shows the static extinction characteristics of an EADFB laser and EADFB SOAs with their SOA lengths of 50, 100, and 150 μm at 45°C. As shown in the figure, there was no SOA-induced degradation in the extinction characteristics. We observed that the inclination of the extinction curve became slightly steeper when ISOA increased from 10 to 25 mA for an SOA length of 50 μm, and we also observed this tendency for 100- and 150-μm-long SOAs with the same current density. There was no serious degradation in the extinction characteristics in any of the above cases.
Figure 10 shows the small signal electrical to optical (E/O) response of an EADFB laser and EADFB SOA with SOA lengths of 50-, 100-, and 150 μm at 45°C. ILD was set at 60 mA. The bias voltage of the EA (Vdc) was set at −2.0 V for each device. The ISOA value for each device was set at a value at which the current densities had the same values. As indicated in the figure, an unwanted peak was observed around 10 GHz and this peak became larger as SOA became longer and ISOA became small. This means that a smaller ISOA will increase the possibility of an unwanted peak occurring. On the other hand, the E/O response of the EADFB SOA with a 50-μm-long SOA did not exhibit an unwanted peak and exhibited a flat response up to around 27 GHz when ISOA was 25 mA. The response of an ISOA of 10 mA also exhibited characteristics that suppressed this peak, so we set an ISOA of 10 mA as the lower limit for the 50-μm-long SOA. The measured 3-dB-down frequency response (f3dB) for the EADFB SOA with an SOA length of 50 μm was about 36 GHz and this is the same as that of an EADFB laser.
To investigate the origin of the peak of the small-signal E/O response at around 10 GHz as shown in Fig. 10, we measured the large-signal response i.e., the intensity waveform launched from the EADFB laser and EADFB SOAs with SOA lengths of 50, 100, and 150 μm at 45°C. As shown in Fig. 11, When the SOA is 150 μm long, the intensity rises and falls greatly in a period of about 0.1 ns, and the rises gently again. We consider this to be caused by the carrier recovery time of about 0.1 ns. This means that when the intensity rises, the carrier in the SOA cavity was quickly consumed, and a new carrier appeared about 0.1 ns later. We consider that this caused the peak at 10 GHz shown in Fig. 10. On the other hand, when the SOA is 50 μm, the difference between the intensity waveform of the EADFB laser and the EADFB SOA are sufficiently reduced. If the carrier density increases, the effective lifetime will be short, and the difference will be further reduced. These results also indicate that the possibility of employing 100- or 150-μm-long SOAs with a larger ISOA while reducing ILD to less than 60 mA may not be effective. This means to suppress the unwanted peak shown in Fig. 10, ISOA values of 50 or 75 mA are needed for 100- and 150-μm-long SOAs, respectively. But from the results in Fig. 6, the power consumptions of 100- and 150-μm-long SOAs with ISOA values of 50 and 75 mA are about 90 and 140 mW, respectively. On the other hand, the power consumption of LD sections with ILD values of 60 and 80 mA are about 76 and 112 mA, respectively. This means that if 100- or 150-μm-long SOAs are driven while suppressing the pattern effect, the power assigned to the LD section will be insufficient. From these results, we choose a 50-μm SOA to achieve our goal of compensating for the output power and chirp of the EA modulator.
We measured the chirp parameter according to a previously described procedure [13]. Figure 12 shows the optical insertion loss of an EA modulator versus the chirp parameter for an EADFB SOA with an SOA length of 50 μm, and an EADFB laser. The optical insertion loss was estimated by subtracting the optical output power when the EA modulator was open from the optical output power when Vdc was applied to the EA modulator. The Vdc dependence of the output power is shown in Fig. 9. As shown in Fig. 12, when the optical loss is small, the chirp parameter of the EADFB laser is reduced. This means that the chirp at the higher optical output power, which corresponds to the 1-level of the optical signal, decreased after traveling through an SOA. This result indicates that the chirp parameter of the EA modulator is compensated for and the limitation caused by the K-K relationship is overcome. This result suggests that clearer eye diagrams will be obtained after an extended distance transmission with the EADFB SOA.
Figure 13 shows 40-Gbit/s eye diagrams for each device. The figure also shows the electrical signal input into the devices, namely a 40-Gbit/s non-return-zero (NRZ), pseudorandom binary sequence (PRBS) of 231-1. The modulation voltage swing (Vpp) and Vdc were 2.9 V and −2.4 V, respectively. As expected, a controlled chirp parameter for the EADFB SOA can provide clear eye diagrams after a 5-km SMF transmission.
The bit-error-rate (BER) performance of back-to-back (BTB) and 2-, 4-, and 5-km SMF transmissions at 40 Gbit/s for the EADFB SOA are shown in Fig. 14. Conventionally, the transmission limit of a 1.55-μm EADFB at 40 Gbit/s is around 2 km, but as a result of the reduced chirp realized with the integrated SOA, we successfully extended the transmission distance to 5 km. These results indicate that when the SOA length is 50 μm, the effect of the pulse pattern shape change described in Fig. 11 is sufficiently suppressed. These results indicate that our concept is valid when the SOA is 50 μm.
Table 1 provides a summary comparison of the characteristics of the EADFB SOA with an SOA length of 50 μm and the EADFB laser. To discuss the feasibility of this approach for integrating a short SOA, we compared the modulated output light power (Pavg) and the power consumption (P) under the ILD and ISOA conditions described in Table 1. The table shows that Pavg increased and P decreased simultaneously with the EADFB SOA. Pavg increased by about 2 dB with almost the same P. These results indicate that this approach using a short SOA will be an attractive technique for overcoming the limitation of the conventional EADFB laser.
Table 1. Typical characteristics of EADFB laser and EADFB SOA
Finally, to investigate the potential for higher speed EADFB SOA operation, we measured eye diagrams for 25-Gbaud/s, 4-pulse amplitude modulation (PAM), which corresponds to a bit rate of 50 Gbit/s with an NRZ format. In [14], we described the measurement setup. Figure 15 shows the eye diagrams of the EADFB laser and the EADFB SOA with an SOA length of 50 μm. Although the difference between the two eye diagrams may be caused by the difference between the optical output power of the EADFB laser and the EADFB SOA, this result clearly indicates that the eye diagrams of the EADFB SOA were not degraded by the pulse pattern shape effect as shown in Fig. 11. These results indicate that our approach for increasing the optical output power and reducing chirp with a low power consumption by using a short SOA is promising in terms of achieving a high-capacity and low-power-consumption light source.
4. Conclusion
We proposed a novel approach that simultaneously increases the output light power and reduces the chirp of an EADFB laser by the monolithic integration of a short cavity SOA. In this approach, the SOA both boosts the output light power and reduces the chirp parameter. To avoid increasing the total power consumption of the device as a result of the injection current of the SOA, we balanced the injection current of DFB laser and that of the SOA as the total output power increased. By using a rate equation as a design guide, the SOA length and ISOA dependence of the calculated gain versus the optical input power with an open bias voltage of EA modulator were investigated. By measuring basic characteristics such as the lasing spectra, extinction ratio, E/O response, static chirp parameter and intensity waveform, we confirmed that an SOA length of 50 μm can realize our concept that compensates for output power and chirp with reduced power consumption compared with a stand-alone EADFB laser. With the device, we achieved a high modulated output light power of over 5 dBm and an extended transmission distance of 5 km at 40 Gbit/s. This approach will provide an attractive way to overcome the limitation of the EADFB laser.
Acknowledgments
We thank Prof. Hiroshi Yasaka of the Research Institute of Electrical Communication, Tohoku University for his valuable advice.
References and links
1. T. Fujisawa, K. Takahata, W. Kobayashi, T. Tadokoro, N. Fujiwara, S. Kanazawa, and F. Kano, “1.3 μm, 50 Gbit/s electroabsorption modulators integrated with DFB laser for beyond 100G parallel LAN applications,” Electron. Lett. 47(12), 708–710 (2011). [CrossRef]
2. K. Naoe, N. Sasada, Y. Sakuma, K. Motoda, T. Kato, M. Akashi, J. Shimizu, T. Kitatani, M. Aoki, M. Okayasu, and K. Uomi, “43-Gbit/s operation of 1.55-μm electro-absorption modulator integrated DFB laser modules for 2-km transmission,” in Proc. ECOC, Th 2.6.4 (2005).
3. W. Kobayashi, M. Arai, T. Yamanaka, N. Fujiwara, T. Fujisawa, T. Tadokoro, K. Tsuzuki, Y. Kondo, and F. Kano, “Design and fabrication of 10-/40-Gbit/s, uncooled electroabsorption modulator integrated DFB laser with butt-joint structure,” J. Lightwave Technol. 28(1), 164–171 (2010). [CrossRef]
4. H. Kim, S. K. Kim, H. Lee, S. Hwang, and Y. Oh, “Novel way to improve the dispersion-limited transmission distance of electroabsorption modulated lasers,” IEEE Photon. Technol. Lett. 18(8), 947–949 (2006). [CrossRef]
5. M. N. Ngo, H. T. Nguyen, C. Gosset, D. Erasme, Q. Deniel, N. Genay, R. Guillamet, N. Lagay, J. Decobert, F. Poingt, and R. Brenot, “Electroabsorption modulated laser integrated with a semiconductor optical amplifier for 100-km 10.3 Gb/s dispersion-free-penalty-free transmission,” J. Lightwave Technol. 31(2), 232–238 (2013). [CrossRef]
6. W. Choi, N. Frateschi, J. Zhang, H. Gebretsadik, R. Jambunathan, A. E. Bond, J. Van Norman, D. Vandegrift, and C. Wanamaker, “Full C-band tunable high fibre output power electroabsorption modulator integrated with semiconductor optical amplifier,” Electron. Lett. 39(17), 1271–1272 (2003). [CrossRef]
7. T. Watanabe, N. Sakaida, H. Yasaka, F. Kano, and M. Koga, “Transmission performance of chirp-controlled signal by using semiconductor optical amplifier,” J. Lightwave Technol. 18(8), 1069–1077 (2000). [CrossRef]
8. K. C. Chang, S. W. Shen, M. C. Hsu, Y. J. Chiu, C. C. Wei, and C. K. Lee, “Negative chirped EAM-SOA for distance insensitive optical OFDM transmission in long reach OFDMA PONs,” in Proceedings of OFC, Tu3H.4 (2014). [CrossRef]
9. J. Mork and A. Mecozzi, “Theory of the ultrafast optical response of active semiconductor waveguide,” J. Opt. Soc. Am. B 13(8), 1803–1816 (1996). [CrossRef]
10. W. Kobayashi, M. Arai, T. Fujisawa, Y. Shibata, T. Sato, T. Ito, K. Hasebe, T. Yamanaka, and H. Sanjoh, “40-Gbit/s EADFB laser integrated with short cavity SOA improving chirp characteristics with low power consumption,” in proc. ECOC, Mo.4.4.5 (2014).
11. H. Yamamoto, M. Hirai, O. Kagaya, K. Nogawa, K. Naoe, N. Sasada, and M. Okayasu, “Compact and low power consumption 1.55-μm electro-absorption modulator integrated DFB-LD TOSA for 10-Gbit/s 40-km transmission,” in Proceedings of OFC, OThT5 (2009).
12. Y. Morita, T. Yamatoya, Y. Hokama, K. Akiyama, R. Makita, N. Yasui, D. Morita, H. Kawahara, and E. Ishimura, “1.3 μm 28 Gb/s EMLs with hybrid waveguide structure for low-power-consumption CFP2 transceivers,”Proceedings of OFC, OTh4H.5 (2013).
13. F. Devaux, Y. Sorel, and J. F. Kerdiles, “Simple measurement of fiber dispersion and of chirp parameter of intensity modulated light emitter,” J. Lightwave Technol. 11(12), 1937–1940 (1993). [CrossRef]
14. T. Fujisawa, Y. Ueda, H. Sanjoh, S. Kanazawa, T. Ito, and W. Kobayashi, “Advantages of EADFB laser for 25 Gbaud/s 4-PAM (50 Gbit/s) modulation and 10 km single-mode fibre transmission,” Electron. Lett. 50(9), 683–685 (2014). [CrossRef]
