Open Access
Add to BibSonomy
Abstract
We demonstrate WDM 256 QAM coherent transmissions with injection locking in the C- and L-bands and compare the transmission performance in the two bands. Although four-wave mixing (FWM) is more significant in an L-band EDFA than in a C-band EDFA, the FWM did not accumulate through the transmission and the FWM components were hidden by the ASE noise level. Since the FWM was weakened by the decorrelation of the WDM signals during the transmission, the transmission performance in the L-band was the same as that in the C-band. The injection locking circuit enabled precise carrier-phase synchronization between a data signal and a local oscillator regardless of the transmission band. By using this circuit, we successfully transmitted 58.2 and 57.6 Tbit/s 256 QAM WDM signals over 160 km with a spectral efficiency of 12 bit/s/Hz in the C- and L-bands, respectively.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The wideband wavelength division multiplexing (WDM) transmission of multi-level digital coherent optical signals is expected to realize a large increase in transmission capacity by virtue of the high spectral efficiency (SE) of the quadrature amplitude modulation (QAM) format. In such a transmission, precise carrier-phase synchronization between transmitted data and a local oscillator (LO) plays a very important role for demodulating a higher-order QAM signal with high accuracy.
Digital carrier-phase estimation is widely employed as a carrier-phase synchronization technique. This technique has already been used to demonstrate 120 Tbit/s 256 QAM WDM coherent transmissions over the C + L-band with an SE of 12 bit/s/Hz [1,2]. However, generally speaking, the accuracy of carrier-phase estimation is insufficient for the precise demodulation of higher order QAM signals. Therefore, the bit error rate (BER) has remained as low as 2 × 10−2 even under a back-to-back condition despite a high optical signal-to-noise ratio (OSNR) [2]. Furthermore, in the digital scheme, the digital signal processor (DSP) load becomes heavier as the QAM multiplicity increases due to the need for higher precision in the phase recovery, which makes the real-time demodulation of higher-order QAM signals difficult.
By contrast, an analog optical carrier-phase synchronization scheme can realize precise phase locking regardless of modulation multiplicity and symbol rate. Injection locking is particularly attractive [3–5] because it enables us to realize precise carrier-phase locking without increasing the DSP load even with a high multiplicity [6], and there is no need to use a narrow linewidth laser as an LO. A real-time 256 QAM transmission has already been realized using the injection-locking scheme [7].
The application of injection locking to a frequency comb-based transmission scheme has been proposed [5,8]. In this scheme, a frequency comb is used as a source for both the transmitter and the LO, and a single pilot tone is shared by many channels, thus reducing the number of sources and simplifying the system configuration. A 44 Tbit/s, 256 QAM transmission over 40 km was realized with an SE of 11.5 bit/s/Hz with this scheme [5]. However, it is generally difficult to obtain a high OSNR for each carrier as many modes are generated from a single comb source. Due to the insufficient OSNR, the transmission distance is limited to 40–50 km. On the other hand, we have developed a precise optical carrier-phase synchronization circuit with injection locking for WDM multi-level QAM transmission [9,10], in which a single LD source is used per channel. With this circuit, we have already demonstrated a 42.3 Tbit/s, 235-channel WDM, 18 Gbaud 64 QAM 160 km transmission in the C-band [9]. Recently, we constructed an injection-locked receiver for the L-band, where we successfully transmitted a 480-channel WDM, 9 Gbaud 128 QAM signal over 160 km [10].
Although the above-mentioned large capacity WDM transmission experiments have already been performed, there has been no detailed comparison of the multi-level QAM transmission performance of the C- and L-bands. To realize high-capacity, long-distance WDM multi-level transmission, it is important to clarify the advantages and disadvantages of C- and L-band transmission and optimize the respective transmission conditions.
In this paper, we describe 485-channel WDM in the C-band / 480-channel WDM in the L-band, 256 QAM transmission with injection locking in the C- and L-bands and compare the performance of ultra multi-level QAM WDM transmissions in the C- and L-bands. Here, by newly employing a Raman amplifier in our WDM transmission system as well as erbium-doped fiber amplifiers (EDFAs), we improved the OSNR and were able to increase the QAM multiplicity to 256. As a result, a WDM 256 QAM signal was successfully transmitted over 160 km with a capacity exceeding 50 Tbit/s and an SE of 12 bit/s/Hz in both bands. Then, we compare the transmission characteristics in both transmission bands including the performance of our injection locking circuit.
2. Experimental setup for 485/480-channel, 256 QAM WDM coherent transmission in C- and L-bands
Figure 1 shows the configuration of our transmission system in the C-band. In the transmitter, 5 LDs were used to generate a channel under the measurement block and 96 LDs (1527.99 nm ∼ 1566.31 nm) were used as a loading dummy channel block. For the channel under measurement block, a tunable external cavity LD (ECLD) with an 8 kHz linewidth [11] was used for the BER measurement, and the other 4 LDs with a 100 kHz linewidth were used as dummy channels. The outputs from these LDs were modulated with 9 Gbaud 256 QAM data. A pilot tone (PT) signal for injection locking was also generated in an arbitrary waveform generator (AWG). The AWG was driven at 63 Gsamples/s with an 8-bit resolution and a 25 GHz analog bandwidth. The data pattern length was 16384 symbols. Here, the power ratio of the QAM and pilot tone signals was optimally set at 10:1, to avoid the S/N degradation of the QAM signals due to the limited ENOB of the D/A and the degradation of the injection locking performance. The bandwidth of the 256 QAM data signal was reduced by using a Nyquist filter with a roll-off factor of 0.05. The PT was attached to each channel, and its frequency was set at 4.92 GHz higher than the carrier frequency of the data signal. For the loading dummy channels, the output from 96 even/odd LDs with a 100 kHz linewidth were phase-modulated by a 10 GHz sinusoidal wave to generate 10 GHz-spaced 480 carriers. These carriers were IQ-modulated in the same manner as the channel under measurement. In this transmission, four dummy channels modulated with identical data were allocated around a test channel in the channel block under measurement to simplify the transmitter setup. In the loading dummy channel block, five carriers generated from a single LD through an optical phase modulator were modulated with identical data. This setup results in a strong correlation between neighboring channels, which may induce excessive nonlinearity during transmission [12]. Therefore, the performance under such conditions may be regarded as the worst case.
We measured the BER characteristics separately for the WDM channels by shifting the wavelength of the tunable ECLD. In a practical system, narrow linewidth LDs are required for all the channels. Compact LD modules with a linewidth narrower than 10 kHz have recently become commercially available, and they are expected to be installed in low-cost, compact transmitters integrated with modulators and DSP circuits in the future.
The channel under measurement and the loading dummy channels were polarization-multiplexed independently, and then combined with a wavelength selective switch (WSS). The channel configuration is shown schematically in Fig. 2, where the channel spacing and guard band were set at 10 GHz and 550 MHz, respectively. Each LD was temperature-controlled, and the oscillation frequency drift was suppressed to within 50 MHz. In this way the overlap of tone signal with neighboring channels can be avoided. A total of 485 channels were accommodated in the 4.85 THz bandwidth, which was limited by the bandwidth of the WSS and WDM coupler. The 485-channel, 144 Gbit/s/ch data were transmitted over two 80-km spans of ultra-large area (ULA) fiber. The transmission power was optimally set at −11 dBm/channel as described later in section 5. The fiber loss for the C + L-band was compensated for by using EDFAs and Raman amplifiers that consisted of 4 pumping LDs (wavelength: 1426, 1446, 1466, 1490 nm) with a total power of approximately 1.2 W. The Raman amplifiers contributed a gain of 12.2 dB to the total gain of 16 dB in each span.
At the receiver, the transmitted signal was divided into 2 paths after an EDFA. In one path, the PT signal was extracted by using a grating type optical tunable filter (OTF) with a bandwidth of 3.7 GHz and a 25 MHz-bandwidth etalon filter (finesse ∼400, FSR = 10 GHz). The extracted PT was injected into a tunable distributed feedback (DFB) laser array [13] with a linewidth of 2.7 MHz in a free running condition. This laser array was used as an LO. The output of the injection-locked LO was frequency downshifted to coincide with the carrier frequency for homodyne detection. In the other path, the QAM data signal was demultiplexed by using an OTF with a bandwidth of 13 GHz, and homodyne detected with the LO signal. Here, the path length difference between the data and LO was adjusted to less than 10 cm to completely match the timing between the two signals to obtain an intermediate frequency (IF) signal with low phase noise [14]. The detected signals were then A/D-converted with a 40-Gsample/s, 16-GHz bandwidth and demodulated offline with a DSP. The DSP carried out polarization demultiplexing, digital back propagation (DBP), and waveform equalization with an adaptive 99-tap finite impulse response (FIR) filter [9]. By using DBP, we compensated for the nonlinear waveform distortion caused by self-phase modulation (SPM) and the cross-phase modulation (XPM) between two orthogonal polarizations. Finally, the QAM signals were demodulated into binary data, and the BER was evaluated from 131-kbit data.
The configuration of the L-band transmission system was almost the same as that used for the C-band described above except for the light sources, EDFAs, WSS, and 90-degree optical hybrid. The linewidth of the L-band LO was 1.9 MHz. The total channel number was reduced to 480 from 485 because of the bandwidth limitations of the WSS and WDM coupler in the L-band.
3. Comparison of C- and L-band injection locking characteristics
First, we compare the carrier-phase locking performance of two injection-locked receivers for the C- and L-bands. Figures 3(a) and 3(b) show the locking range of LO at 1546 and 1589 nm for the C- and L-bands, respectively. In the high injection power region over −2.5 dBm, the injection locking became unstable, where the injection-locked LO began to be intensity modulated. Therefore, the injection power was set at −5 dBm to obtain a wideband locking range while maintaining a stable locking condition. Here, the locking range was 8.2 GHz for the C-band. In the L-band, we obtained a locking range of 7 GHz, which was almost the same as that for the C-band.
To evaluate the carrier-phase locking performance, we measured the 4.92 GHz IF signal of the beat between the PT and the injection-locked LO under a back-to-back condition at an injection power of −5 dBm. Figures 4(a) and 4(b) show the IF spectra with injection locking within a 2 MHz span for the C- and L-bands, respectively. The resolution bandwidth was 1 kHz. The corresponding single-sideband (SSB) phase noise spectra are shown in Figs. 4(c) and 4(d). A low noise IF signal was obtained with a signal-to-noise ratio (SNR) of approximately 80 dB. The phase noise, which was evaluated by integrating the SSB noise power spectrum from 10 Hz to 1 MHz, were only 0.21 and 0.22 degrees in the C- and L-bands, respectively. Figures 5(a) and 5(b) show the wavelength dependence of the SSB phase noise for the C-band (1527 nm ∼ 1567 nm) and L-band (1570 nm ∼ 1610 nm), respectively. A phase noise variance of less than 0.23 degrees was obtained over the entire C-band range. The phase noise for the L-band was less than 0.25 degrees. These values were sufficiently smaller than the angle of 2.0 degrees between adjacent symbols in the 256 QAM format. This result indicates that our injection-locked coherent receivers have highly precise phase synchronization characteristics in both the C- and L-bands.
Fig. 4. IF spectra between PT and injection-locked LO in a 2 MHz span and SSB phase noise spectra (10 Hz ∼ 1 MHz). (a), (c) in the C-band, and (b), (d) in the L-band.
Fig. 5. Wavelength dependences of SSB phase noise of the IF spectrum. (a) in the C-band and (b) in the L-band.
The phase noise variance of an IF signal generated by injection locking is inversely proportional to the locking range of an LO, which strongly depends on the Q value of the laser cavity and is independent of oscillation wavelength [15]. As the configurations of the DFB LDs used as LOs for the C- and L-bands were the same, these LDs would have the same Q value. Therefore, their locking range characteristics were almost the same as shown in Fig. 3. In general, the phase noise can be reduced by increasing the locking range [15], which indicates that a wider locking range is beneficial for highly precise phase synchronization. The phase noise variance is proportional to the linewidth of the transmitter laser [15]. It is therefore important to use a narrow linewidth LD as a transmitter to realize low-noise injection locking. In our transmission system, the linewidths of the transmitter lasers in C- and L-bands were the same. Therefore, the same IF signal characteristics were obtained regardless of the transmission band.
4. Comparison of EDFA and Raman amplifier characteristics in C- and L-bands
Next, we compare the characteristics of EDFA and Raman amplifiers in the C- and L-bands, in terms of nonlinearity, gain, and noise characteristics.
In general, an L-band EDFA requires an EDF that is longer than that needed in a C-band EDFA to achieve the same gain. Moreover, the effective core area (Aeff) of the EDF is smaller than that of standard single mode fiber. Therefore, an L-band EDFA is more susceptible to nonlinear effects than a C-band EDFA [16]. In this experiment, the EDF length of the L-band EDFA was 32 m, which was 2.2 times that of the C-band EDFA (14.5 m). The Aeff values of the C- and L-band EDF were 27 and 30 µm2, respectively, namely they were almost the same for both bands. In this experiment, dummy channels were obtained from a single IQ modulator, which resulted in a strong data pattern correlation among the channels. These conditions easily cause four-wave mixing (FWM) during propagation inside an EDFA.
We used five QAM data channels to measure the FWM components of the transmitter and receiver generated in the EDFA under a back-to-back condition using the setup shown in Fig. 1. Here, the frequency separation was set at 50 GHz for edge channels and 100 GHz for center channels to allow us to observe the FWM components. The optical power per channel was set at the same value as in the 480 ch-WDM transmission experiment. The output powers of the EDFAs in the transmitter and receiver were 18 and 13 dBm, respectively. Figures 6(a) and 6(b) show the optical spectra of the five channels in the C- and L-bands, respectively, measured at the output of the EDFA before OTF. The resolution bandwidth was set at 0.1 nm. In the C-band spectrum shown in Fig. 6(a), very small humps can be seen on both sides of the center channel. These are the components produced by FWM. Here, the ratio of the data signal power Ps to the FWM component power PFWM (PS /PFWM) was 40.7 dB for the C-band, and the OSNR was 41 dB. For the C-band, the FWM and amplified spontaneous emission (ASE) are at almost the same level, hence the influence of the FWM components can be negligible.
Fig. 6. Optical spectra of 5 ch-WDM signal measured at the EDFA output before OTF–1 in Fig. 1 under the back-to-back condition. For (a) C-band (b) L-band.
Fig. 7. Optical spectra of 485/480 ch-WDM signal after 160 km transmission. For (a) C-band and (b) L-band.
On the other hand, the FWM components are clearly observed in the L-band spectrum shown in Fig. 6(b). Although the OSNR was 41 dB, the PS /PFWM value was 38.6 dB, which was 2.1 dB lower than that of the C-band. This FWM component is generated by a strongly correlated pair of channels. As described in section 2, five carriers are modulated with the same data when generating a 480 ch WDM signal, which causes FWM among a number of different channel combinations. As a result, when the 480 ch WDM signal is amplified, FWM caused by 10 pairs of channels can overlap the channel under measurement. This indicates that an FWM component that is 10 times larger than that in Fig. 6(b) can affect the demodulation performance. On the other hand, after a 160 km transmission, the FWM level remained the same. This is because the correlation among dummy channels is weakened due to the group delay between the channels in the transmission fiber. This indicates that the FWM occurred mainly in the EDFA at the transmitter, where a WDM signal with strong channel correlation is amplified.
Fig. 8. Enlarged view of 480 ch-WDM signal in the L-band after 160 km transmission when neighboring dummy channels were eliminated for OSNR measurement.
Then, we compared the gain and noise characteristics. Figures 7(a) and 7(b) show the optical spectra of 485/480-channel WDM 256 QAM signals after 160 km transmissions for the C- and L-bands, respectively. For the C-band, the flatness of the WDM spectrum was 1.8 dB before transmission, but after a 160 km transmission, it was degraded to 6.2 dB due to the wavelength dependences of the gains of the EDFAs and Raman amplifiers. For the L-band, the flatness values before and after transmission were 0.8 and 4.4 dB, respectively, which were lower than those of the C-band due to the better gain flatness of the EDFAs and Raman amplifiers in the L-band.
Fig. 9. Wavelength dependences of OSNR of transmitted 485/480 channel-WDM 256 QAM data signal. For (a) C-band and (b) L-band.
Figure 8 shows enlarged view of 480 ch-WDM signal in the L-band after 160 km transmission when two neighboring dummy channels (100 GHz) were eliminated for OSNR measurement. In the region where the signals are turned off, the noise component is flat. This indicates that the ASE noise is dominant in this region and the FWM component is below the ASE noise level. We estimated the OSNR from the difference between the signal level in the signal spectrum and the ASE level in the region where the signal is off. The measured OSNR values are shown in Fig. 9(a) for the C-band and 9(b) for the L-band measured after a 160 km transmission. The OSNRs were 27.8∼33.9 dB and 30.8∼32.8 dB for C- and L-bands, respectively. These values were improved by an average of ∼ 4 dB compared with transmission using only an EDFA. The OSNR variation in the C-band was 6.2 dB, while it was only 2 dB in the L-band. Such a large difference is due to the gain flatness of the EDFA and Raman amplifiers as shown in Fig. 7(a) and the noise figure characteristics of the Raman amplifier in which OSNRs of shorter wavelength regions tend to be relatively low [17].
5. Comparison of C- and L-band transmission characteristics
As the next step, we compared the demodulation characteristics under a back-to-back condition. Figures 10(a) and 10(b) show the BER characteristics for a 9 Gbaud 256 QAM signal as a function of the OSNR for a single-channel case and a WDM case in the C-band and L-band, respectively. Theoretical BER curves are also shown. It can be seen that there is an OSNR penalty between the single-channel BER and the theoretical BER curves, which was 2.9 dB for the C-band and 3.2 dB for the L-band at a BER of 1 × 10−2. Furthermore, there is an error floor in these BER characteristics. This OSNR penalty and error floor would be caused by imperfect waveform distortion compensation using a time-domain 99-tap FIR filter, the limited ENOBs of the D/A and A/D and nonlinearities in optical modulators and electrical components. An OSNR penalty of 2.1 dB was observed between single-channel and WDM cases at a BER of 2 × 10−3 for the L-band as shown in Fig. 10(b), while it was negligibly small for the C-band as shown in Fig. 10(a). This difference was caused by crosstalk due to the FWM between dummy channels that occurred inside the L-band EDFAs as described in section 4.
Fig. 10. BER characteristics for back-to-back condition. (a) C-band, ch. 243 at 1546.92 nm, (b) L-band, ch. 243 at 1589.57 nm.
Figure 11 shows the error vector magnitude (EVM) of a demodulated 9 Gbaud 256 QAM signal as a function of the transmission power measured at the center wavelength of each band (ch. 243) after a 160 km transmission. Here, we adopted a standard DBP that can compensate for XPM between two orthogonal polarizations based on the Manakov equation [18]. Since the transmission performance is dominantly determined by the XPM between WDM channels, the improvement in the demodulation performance is insufficient with DBP alone, where the BER improvement was only 10% error reduction after DBP. Based on this result, we set the fiber launch power at −11 dBm/channel (15.9 dBm/485 channels for the C-band, 15.8 dBm/480 channels for the L-band). At launch powers below −11 dBm/channel, the EVM was degraded because OSNR degradation was attributed to increase of the ASE noise, while it also was degraded due to fiber nonlinearity at a higher launch power. Although the optimum launch power was 1 dBm in a single channel 256 QAM transmission [6], it was reduced to −11 dBm for WDM transmission to avoid inter-WDM channel XPM.
Since the wavelength dependence of Aeff is small, with only a 5.3% difference between the C- and L-bands for ULA fiber [19], the optimum launch power derived from Fig. 11 was the same for both bands. This result indicates that the influence of nonlinearity in the transmission fiber is almost the same for the C- and L-bands.
Then, we measured the BER characteristics for 52 channels that covered the entire C- and L-band after transmission. Figure 12(a) shows the BER characteristics for the C-band. The BER characteristic is degraded in the short wavelength region because of the OSNR degradation in this wavelength region as shown in Fig. 9(a). However, the BERs of all the measured channels were below the forward error correction (FEC) threshold with a 20% overhead (2 × 10−2). In ch. 23 at 1529.55 nm, the BER was 1.79×10−2 with an OSNR of 27.7 dB. This BER was the worst in the present transmission experiment. In ch. 433 at 1562.23 nm, we obtained a BER of 4.83×10−3 with an OSNR of 34.4 dB, which was the best BER. Here, the BER of the lowest OSNR channel is still below the FEC threshold. Therefore, we expect that a BER below the FEC threshold can be achieved for all the other channels because their OSNRs are higher. Figure 12(b) shows the constellation of ch. 23 after a 160 km transmission. The corresponding EVM was 3.28%. In this transmission, a potential 69.84 Tbit/s data capacity was transmitted within an optical bandwidth of 4.85 THz over the C-band, resulting in an SE of 12 bit/s/Hz by taking a 20% FEC overhead into account. The net data capacity reached as high as 58.2 Tbit/s.
Fig. 12. 160 km transmission results in the C-band. (a) BER characteristics of 52 channels. (b) Constellation for 9 Gbaud 256 QAM signal.
Figure 13(a) shows the wavelength dependence of the BERs for L-band transmission. All 52 measured BERs were below the FEC threshold with a 20% overhead. The BERs were almost uniform regardless of wavelength. This may be attributed to the fact that there is less OSNR variation in the L-band as shown in Fig. 9(b). In ch. 3 at 1569.59 nm, we obtained a BER of 1.91×10−2 with an OSNR of 30.8 dB, which was the worst BER. In ch. 243 at 1589.57 nm, the BER was 7.34×10−3 with an OSNR of 32.1 dB, which was the best BER. The constellation of ch. 3 is shown in Fig. 13(b). Its EVM was 3.29%. We successfully transmitted data at 69.12 Tbit/s within an optical bandwidth of 4.80 THz over the L-band. The SE reached 12 bit/s/Hz by assuming a 20% FEC overhead. The net data capacity was 57.6 Tbit/s.
Fig. 13. 160 km transmission results in the L-band. (a) BER characteristics of 52 channels. (b) Constellation for 9 Gbaud 256 QAM signal.
We obtained BERs below the FEC threshold of 2 × 10−2 for both C- and L-band transmissions. The BERs in the L-band were almost the same as those in C-band even though the L-band has a relatively higher OSNR. This may be attributed to the additional waveform distortions caused by fiber nonlinearity such as FWM in the L-band EDFA at the transmitter as described in section 4.
6. Conclusion
We demonstrated 485/480-channel WDM 9 Gbaud 256 QAM 160 km transmission in the C- and L-bands with an injection locking technique and compared the performance in the two bands. By newly introducing a Raman amplifier, we were able to improve the OSNR of transmitted 256 QAM data signals. With the present injection-locked coherent receiver, we achieved low-noise carrier-phase synchronization over the C- and L-band range. We showed that precise carrier-phase locking characteristics can be obtained regardless of the transmission band by using an LO with a low Q value and a narrow linewidth transmitter laser. As a result, we achieved 58.2 and 57.6 Tbit/s transmissions for the C- and L-bands, respectively. The SE reached as high as 12 bit/s/Hz. These represent the highest capacity yet obtained in each band by employing injection locking.
Although FWM is more significant in an L-band EDFA than in a C-band EDFA, the FWM did not accumulate through the transmission and the FWM components were concealed by the ASE noise level. Since the FWM was weakened by the decorrelation of the WDM signals during the transmission, there is no obvious difference between the nonlinear impairments in the transmission fiber for the C- and L-bands. We believe that these results are important in relation to the practical application of large-capacity ultra-multi-level WDM coherent transmission.
Funding
Japan Society for the Promotion of Science (JP19J12822).
Disclosures
The authors declare no conflicts of interest.
References
1. M. Ionescu, L. Galdino, A. Edwards, J. James, W. Pelouch, E. Sillekens, D. Semrau, D. Lavery, R. I. Killey, S. Barnes, P. Bayvel, and S. Desbruslais., “91 nm C + L hybrid distributed Raman–erbium-doped fibre amplifier for high capacity subsea transmission,” in Proc. European Conference on Optical Communication 2018, paper Mo4G.
2. Y. Yu, L. Jin, Y. Lu, Y. Huang, and L. Li, “400 × 377.6 Gb/s C + L band seamless transmission over 200 km with Raman amplifier using PDM–256QAM,” in Proc. European Conference on Optical Communication 2018, paper Th2.49.
3. S. Adhikari, S. Sygletos, A. D. Ellis, B. Inan, S. L. Jansen, and W. Rosenkranz, “Enhanced self-coherent OFDM by the use of injection locked laser,” in Proc. Optical Fiber Communication Conference 2012, paper JW2A.
4. Z. Liu, J. Y. Kim, D. S. Wu, D. J. Richardson, and R. Slavík, “Homodyne OFDM with optical injection locking for carrier recovery,” J. Lightwave Technol. 33(1), 34–41 (2015). [CrossRef]
5. M. Mazur, J. Schröder, A. L. Riesgo, T. Yoshida, M. Karlsson, and P. A. Andrekson, “11.5 bits/s/Hz PM–256QAM comb-based superchannel transmission by combining optical and digital pilots,” in Proc. Optical Fiber Communication Conference 2018, paper M1G.2.
6. Y. Wang, K. Kasai, M. Yoshida, and M. Nakazawa, “320 Gbit/s, 20 Gsymbol/s 256 QAM coherent transmission over 160 km by using injection-locked local oscillator,” Opt. Express 24(19), 22088–22096 (2016). [CrossRef]
7. M. Yoshida, T. Kan, K. Kasai, T. Hirooka, and M. Nakazawa, “Real-time 256 QAM bi-directional coherent transmission in next generation mobile fronthaul,” in Proc. OptoElectronics and Communications Conference 2019, Paper TuA2–3.
8. M. Mazur, A. L. Riesgo, J. Schröder, P. A. Andrekson, and M. Karlsson, “High spectral efficiency PM–128QAM comb-based superchannel transmission enabled by a single shared optical pilot tone,” J. Lightwave Technol. 36(6), 1318–1325 (2018). [CrossRef]
9. T. Kan, K. Kasai, M. Yoshida, and M. Nakazawa, “42.3 Tbit/s, 18 Gbaud 64 QAM WDM coherent transmission over 160 km in the C-band using an injection-locked homodyne receiver with a spectral efficiency of 9 bit/s/Hz,” Opt. Express 25(19), 22726–22737 (2017). [CrossRef]
10. T. Kan, K. Kasai, M. Yoshida, T. Hirooka, and M. Nakazawa, “50.4 Tbit/s, 128 QAM L-band WDM injection-locked coherent transmission over 160 km with spectral efficiency of 10.5 bit/s/Hz,” in Proc. OptoElectronics and Communications Conference 2018, paper 5B3–3.
11. K. Kasai, M. Nakazawa, Y. Tomomatsu, and T. Endo, “1.5µm, mode-hop-free full C-band wavelength tunable laser diode with a linewidth of 8 kHz and a RIN of −130 dB/Hz and its extension to the L-band,” Opt. Express 25(18), 22113–22124 (2017). [CrossRef]
12. R. Dar, S. Chandrasekhar, A. H. Gnauck, B. Li, J. Cho, E. C. Burrows, and P. J. Winzer, “Winzer Impact of WDM channel correlations on nonlinear transmission,” in Proc. European Conference on Optical Communication 2016, Paper W.1.D.2.
13. H. Ishii, K. Kasaya, and H. Oohashi, “Spectral linewidth reduction in widely wavelength tunable DFB laser array,” IEEE J. Sel. Top. Quantum Electron. 15(3), 514–520 (2009). [CrossRef]
14. Y. Wang, K. Kasai, M. Yoshida, and M. Nakazawa, “Single-carrier 216 Gbit/s, 12 Gsymbol/s 512 QAM coherent transmission over 160 km with injection-locked homodyne detection,” in Proc. Optical Fiber Communication Conference 2017, paper Tu2E.1.
15. A. C. Bordonalli, C. Walton, and A. J. Seeds, “High-performance phase locking of wide linewidth semiconductor lasers by combined use of optical injection locking and optical phase-lock loop,” J. Lightwave Technol. 17(2), 328–342 (1999). [CrossRef]
16. H. Ono and M. Yamada, “Four-wave mixing crosstalk measurement in a highly doped L-band erbium-doped fiber amplifier by using half of the signal channels,” J. Lightwave Technol. 26(14), 2175–2183 (2008). [CrossRef]
17. J. Bromage, “Raman amplification for fiber communications systems,” J. Lightwave Technol. 22(1), 79–93 (2004). [CrossRef]
18. C. Paré, A. Villeneuve, P.-A. Bélanger, and N. J. Doran, “Compensating for dispersion and the nonlinear Kerr effect without phase conjugation,” Opt. Lett. 21(7), 459–461 (1996). [CrossRef]
19. G. P. Agrawal, Nonlinear Fiber Optics (Academic, 1989) Chap.2.
