In an optical filter based VSB-DD transmission system, semiconductor optical amplifier (SOA) is a promising option to enhance system optical power margin. While, in practical system, the low input saturation power makes the SOA-amplified signal susceptible to the pattern effect, which causes a considerable spectral broadening, thereby influencing the design of VSB filter. In this paper, the relationship between SOA-induced pattern effect and the requirements of the VSB filter is systematically investigated. Firstly, qualitative analysis is given and upper sideband (USB) is proved better than lower sideband (LSB) owing to the suppression of SOA-induced pattern effect. Then, 56Gbps IM/DD PAM4 transmission is experimentally conducted. With respective optimal filter configuration, performance of USB signal is superior to LSB signal in all cases. Results show that USB signal has 1dB sensitivity superiority to LSB signal for 56Gb/s PAM4 after 40km transmission. And in 80km case, only by using USB signal, can HD-FEC limit (3.8 × 10−3) be achieved. Also, we study requirements on other filter parameters, including redundant bandwidth and filter steepness.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Recently, various candidate optical transmission schemes have been presented and discussed to achieve higher data rate as well as longer distance to cater to the explosively increasing bandwidth demand of diverse services . Among these solutions, the intensity modulation/direct detection (IM/DD) with four-level pulse amplitude modulation (PAM4) format is a cost-effective and actively-discussed candidate [2,3]. On one hand, the IM/DD system has simple transceiver structure compared with coherent systems. On the other hand, PAM4 has a reduced digital signal processing (DSP) complexity and relaxed linearity requirements on electrical and optical devices, compared with other high-order modulation formats, such as discrete multi-tone (DMT) and carrier-less amplitude and phase modulation (CAP) [4–6]. However, the double sideband (DSB) feature makes the intensity modulated signal suffer from the power fading caused by chromatic dispersion, which limits the achievable end-to-end bandwidth and transmission distance. To deal with this issue, single sideband (SSB) based transmission scheme is proposed. To generate SSB signal, IQ modulator scheme and filtering scheme are the common methods. For IQ modulator-based scheme, where the in-phase and quadrature beams are respectively injected into the original signal and its Hilbert transform counterpart [7–9]. Although an ideal SSB signal can be directly obtained in this way, complex computation of the Hilbert transform is required, as well as an additional DAC with relatively high resolution. The other method is by filtering an intensity modulated signal to obtain a vestigial sideband (VSB) signal. In this way, cost-effective intensity modulators such as single drive Mach-Zehnder modulator (MZM) or electro-absorption modulated laser (EML), and DAC with 2-bit resolution can be adopted [10–14].
In practical VSB IM/DD systems, optical amplifier is indispensable to compensate the power loss from both long-distance fiber and VSB filter. Previously, erbium-doped fiber amplifiers (EDFAs) are primarily used. But there is seldom study focusing on semiconductor optical amplifiers (SOAs) in VSB-PAM4 system, which is a promising component in future IM/DD based high-data-rate transmission system due to its low cost and easy integration [15–17]. However, an undesired drawback of SOAs is that the low input saturation power makes the SOA amplified signals easily suffer from a considerable spectral broadening when the input peak power exceeds the input saturation power limit, which is referred to as pattern effect . The broadening spectrum causes an asymmetric spectrum, thus influencing the selection of upper or lower sideband as well as the VSB filter design.
In this work, we systematically investigate the relationship between SOA-induced pattern effect and the requirements on the VSB filter. Theoretical analysis indicates that upper sideband (USB) has better performance than lower sideband (LSB) owing to the inhibition of SOA-induced pattern effect. According to the experimental results of 56Gbps VSB IM/DD PAM4 transmission, USB signal has 1dB sensitivity superiority to LSB signal for 56Gb/s PAM4 after 40km transmission. And in 80km case, only by using USB signal, can HD-FEC limit (3.8 × 10−3) be achieved. Also, optimal redundant bandwidth and requirement on steepness are discussed in detail. Results show that for longer distance, smaller roll-off factor is needed.
2. Experimental setup
To explore the requirements of the VSB filter, we conduct the experiment as depicted in Fig. 1. The continuous optical wave (1550nm, 193.548THz) from a laser followed by a polarization controller (PC) is injected into a single-drive MZM (Fujitsu FTM7939EKL). 56Gb/s baseband electrical PAM4 signal with a 215-bit pseudo random bit sequence (PRBS) is generated by arbitrary waveform generator (AWG), featuring a 64-GSa/s sampling rate and 8-bit resolution. Note that, no pre-processing is applied at the transmitter side such that the AWG used here can be displaced by a cost-effective DAC with 2-bit resolution. Before applying the PAM4 signal to MZM, the peak to peak voltage (Vpp) is firstly optimized to obtain a good balance between carrier to signal power ratio (CSPR) and extinction ratio (ER). The MZM is biased at the quadrature point with an output power of 4dBm. After standard single mode fiber (SSMF) transmission, the signal is amplified by a SOA (BOA1004PXS) and then injected into an optical filter (Finisar WaveShaper 4000S), which is employed as a VSB filter as well as an amplified spontaneous emission (ASE) noise filter. Next, a 95:5 optical splitter is used to tap out 5% of the optical power to an optical spectrum analyser (OSA) with a resolution of 150MHz for spectrum monitoring. After adjusting received optical power (ROP) by a variable optical attenuator (VOA), the optical signal is detected by into a photodiode integrated with trans-impedance amplifier (PIN-TIA) with a 3-dB bandwidth of 33GHz. The electrical signal is captured by a digital storage oscilloscope (DSO) featuring a sampling rate of 120GS/s with 45GHz bandwidth. Then, the captured digital signal is feed forward equalized (FFE) or Volterra equalized to compensate the distortions from chromatic dispersion and signal-to-signal beat interference (SSBI) . Note that, Kramers-Korning receiver based signal compensation can also be adopted to mitigate SSBI problem . And according to the results in , KK and volterra-based equalization can lead to similar achievable performance. Finally, bit error rate (BER) is calculated.
3. Pattern effect of SOA and its impact on VSB-PAM4 transmission
Figure 2 depicts the gain as a function of input power of the adopted SOA at 280mA, 25 °C. As shown, the input saturation power (~-7.4 dBm) determines the boundary between linear and nonlinear regions. When the peak power of input PAM4 signal exceeds the input saturation power, additional pattern effect will occur, which is because gain saturation induces self-phase modulation (SPM).
From the perspective of frequency domain, the pattern effect will result in frequency chirp. According to the analysis in , the chirp of the output pulse can be given byFig. 3(a). When SOA works in linear region, the optical spectrum is almost symmetric, which indicates that the PAM4 signal experiences negligible frequency chirp in this case. In contrast, when SOA works in nonlinear region, the optical spectrum is asymmetric and the LSB exhibits a higher power density than the USB. In VSB PAM4 systems, such an asymmetric optical spectrum can influence the selection of USB or LSB. Particularly, when the USB is selected, the undesired frequency chirp contained in LSB can be naturally eliminated, thus a better BER performance is obtained in comparison to the selection of LSB.
As for time domain, the pattern effect distorts the pulse shape of the transmitted signal. Specifically, the instantaneous SOA gain is given byFigures 3(b) and 3(c) show the simulated pulse shapes and eye diagrams of DSB PAM4 pulse in linear region and nonlinear region, where the pattern of ”01013021123230” is taken as an example. As shown in Fig. 3(b), the leading edge is sharper than the trailing edge of each pulse, which is because the leading edge has a larger gain than the trailing edge, according to Eqs. (3) and (4). At the meantime, from Eq. (4), we know that a higher will enlarge the gain difference between the leading edge and the trailing edge. Thus, it can be inferred that the distortion of the pulse shape is intensity-dependent, and the distortion for higher levels of PAM4 signal is more serious. In addition, as shown in Fig. 3(c), the unequally-spaced levels also reveals the intensity-dependent gain in nonlinear region. To conclude, when SOA wokrs in nonlinear region, the nonlinear distortions of signal include non-uniform gain and intensity-dependent noise power.
4. Experimental results and discussion
4.1 Comparison of USB and LSB signal
To fairly compare the USB and LSB signals, we firstly tune the optical filter to optimize their performances respectively. We find that, for both USB and LSB signals, the optimal BER performance is achieved when the 3dB cut-off frequency of the rectangular filter is exactly the frequency of optical carrier signal, regardless of the roll-off factor. For convenience, we define the filter redundant bandwidth as the frequency difference between the lower (upper) 3dB cut-off frequency of the USB (LSB) filter and the optical carrier frequency. And a negative redundant bandwidth means that the overlap of VSB filter and optical signal is smaller than the optimal case . Note that the influence of different redundant bandwidth values will be discussed detailedly in section 4.2. Optical spectra of USB and LSB signal at BtB case are shown in Fig. 4(a), where a roll-off factor of 0 is adopted. It can be seen that the power density of LSB is slightly higher than that of USB. According to the analysis in Section 3, such power density difference is due to the undesired frequency chirp caused by the pattern effect. The histograms of USB and LSB after FFE equalization are also shown in Fig. 4(b). It is observed that the four levels of USB signal slightly distribute unevenly due to the non-flat gain. In addition, for LSB signal, the upper levels are more noised.
Then, performances of USB and LSB after different transmission distances are compared based on the redundant bandwidth of 0GHz and roll-off factor of 0. The input optical powers of SOA for BtB, 40km and 80km cases are respectively −8dBm, −8dBm and −12dBm, of which the peak power (instantaneous power when transmitting level 3) (about −5dBm, −5dBm and −9dBm) exceeds the input saturation power, thus causes a certain amount of pattern effect. The corresponding eyediagrams as well as BER performances are presented in Fig. 5. Overall, the USB signal is better than LSB signal in various distance cases by untilizing either FFE equalization or Volterra equalization. Due to the uneven gain in nonlinear region, the levels of PAM4 signal are not equally-spaced for both LSB and USB signals. But for LSB signal, the upper levels are more noised, which is because the pattern effect caused noise mainly concentrates on LSB signal. It also can be seen that Volterra equalizer can partly settle the issue of uneven gain, but cannot eliminate the noise induced by pattern effect.
According to the BER curves in Fig. 5(b), in BtB case and with FFE equalization, the HD-FEC BER limit (3.8 × 10−3) can be satisfied at −8.8dBm for USB signal and −7.4dBm for LSB signal. With Volterra equalization, sensitivity of both USB and LSB signals can be reduced to −9.4dBm and −8.2dBm due to the mitigation of SSBI and other nonlinear distortions. In 40km case, the performance becomes worse compared with BtB case due to chromatic dispersion and fiber nonlinearity. Nevertheless, superiority of USB signal compared with LSB signal is still observed that 1dB sensitivity superiority can be obtained in the USB signal. In 80km case, the performance further worsens due to the increased chromatic dispersion. In this case, the HD-FEC limit can be achived only by using USB signal and Volterra equalization.
4.2 Analysis on the redundant bandwidth and roll-off factor
Next, the influence of redundant bandwidth on the performance is analyzed. The BER results in BtB case with various redundant bandwidths are plotted in Fig. 6(a). Note that the input optical power of SOA here is −8dBm, which is exactly in the nonlinear region. Since the signal experiences zero chromatic dispersion and no fiber nonlinearity in BtB case, the distortion mainly comes from SOA induced pattern effect. Specifically, when is −2.5GHz, BER performances of both USB and LSB are poor due to the suppressed optical carrier which vastly reduces the CSPR. It decreases rapidly when goes up to 0GHz. Interestingly, if we further increase the redundant bandwidth of LSB signal to 2.5GHz, the BER deterioration is observed again. It can be explained by the insufficient suppression of the vestigial sideband. Also, the variation of redundant bandwidth can influence the amount out-of-band ASE noise, thus slightly affect the BER performance of signal. But this variation can be neglected, comparing with factors such as CSPR and suppression degree of sideband.
The BER curves of USB versus redundant bandwidth at BtB, 40km and 80km with Volterra equalization are plotted in Fig. 6(b). Poor BER can be observed for both 40km and 80km transmission when is −2.5GHz due to the low power of optical carrier. Then, at 0GHz, relatively good BER performance is obtained thanks to the tolerance of VSB signal to power fading effect. The BER even performs better when the is 2.5GHz at BtB and over 40-km SSMF, but gets worse after 80km transmission. Furthermore, when the redundant bandwidth further increases, performance degradation is observed. This is because the insufficient suppression of vestigial sideband makes the optical signal partly convert bake to DSB signal, thereby suffering from power fading effect. One can also easily find that, such degradation is more serious in 80km cases, due to the larger chromatic disperison amount and more serious power fading distortion.
In addition, the influence of filter steepness is discussed. The profiles of VSB filter with different roll-off factors with Volterra equalization are plotted in Fig. 7(a). And the BER results in BtB case under various roll-off factor are shown in Fig. 7(b). There is a battle between signal deterioration caused by pattern effect and peformance improvement resulted from increased signal power, and the latter is domainant for USB signal while the former is domainant for LSB signal in the roll-off factor range of 0.1 to 0.5. Thus, BER of USB decreases as that of LSB increases. Conversely, when further increasing it from 0.5 to 0.7, because nonlinear distortions and useful signal power are repectively the leading parts in USB and LSB, BER of USB increases as that of LSB decreases. Meanwhile, the BER versus roll-off factor under BtB, 40km and 80km cases are depicted in Fig. 7(c). Because of no chromatic dispersion induced power fading, increased roll-off factor doens’t incur performance degradation in BtB case. Different from BtB case, BER results of both 40km and 80km worsen with the increase of roll-off factor. It is because the distotion of power fading is more serious in longer distance, thus steeper filters are required.
The relationship between SOA-induced pattern effect and the requirements of the VSB filter is systematically investigated. SOA’s saturation power can be low and when it is in the saturation region, SOA amplified signals suffer from a considerable spectral broadening, leading to an asymmetric spectrum and pulse shape. According to the theoretical analysis, USB signal has better performance than LSB signal because SOA-induced pattern effect mainly concentrates in the LSB. Experiment of 56Gbps PAM4 transmission is conducted to verify the analysis. We find that USB signal has 1dB sensitivity superiority to LSB signal after 40km transmission. And in 80km case, only by using USB signal, can HD-FEC limit be achieved. Moreover, results show that the optimal 3dB cut-off frequency of the rectangular filter is exactly the frequency of optical carrier signal. Once taking roll-off factors into consideration, for longer distance, smaller roll-off factor is needed.
National Natural Science Foundation of China (NSFC) (61431009, 61521062).
1. P. Miguelez, “What applications are driving higher capacity in access,” in Optical Fiber communications conference (OFC) (2018), paper M2B.1.
2. K. Zhang, Q. Zhuge, H. Xin, Z. Xing, M. Xiang, S. Fan, L. Yi, W. Hu, and D. Plant, “Demonstration of 50Gb/s/λ symmetric PAM4 TDM-PON with 10G-class optics and DSP-free ONUs in the O-band,” in Optical Fiber communications conference (OFC) (2018), paper M1B.5.
3. K. Zhang, Q. Zhuge, H. Xin, M. Morsy-Osman, E. El-Fiky, L. Yi, W. Hu, and D. V. Plant, “Intensity directed equalizer for the mitigation of DML chirp induced distortion in dispersion-unmanaged C-band PAM transmission,” Opt. Express 25(23), 28123–28135 (2017). [CrossRef]
4. C. Kottke, C. Caspar, V. Jungnickel, R. Freund, M. Agustin, J. R. Kropp, and N. Ledentsov, “High-speed DMT and VSCEL-based MMF transmission using pre-distortion,” J. Lightwave Technol. 36(2), 168–174 (2018).
5. K. Zhong, X. Zhou, T. Gui, L. Tao, Y. Gao, W. Chen, J. Man, L. Zeng, A. P. T. Lau, and C. Lu, “Experimental study of PAM-4, CAP-16, and DMT for 100 Gb/s short reach optical transmission systems,” Opt. Express 23(2), 1176–1189 (2015). [CrossRef] [PubMed]
6. J. Shi, J. Zhang, X. Li, N. Chi, G. Chang, and J. Yu, “112 Gb/s/lamda CAP signals transmission over 480km in IM-DD system,” in Optical Fiber communications conference (OFC) (2018), paper W1J.5.
7. M. Zhu, J. Zhang, H. Ying, X. Li, M. Luo, Y. Song, F. Li, X. Huang, X. Yi, and K. Qiu, 56-Gb/s optical SSB PAM-4 transmission over 800-km SSMF using DDMZM transmitter and simplified direct detection Kramers-Kronig receiver,” in Optical Fiber communications conference (OFC) (2018), paper M2C.5.
8. Z. Xu, M. O’Sullivan, and R. Hui, “Spectral-efficient OOFDM system using compatible SSB modulation with a simple dual-electrode MZM,” in Optical Fiber communications conference (OFC) (2010), paper OMR2.
9. S. Fan, Q. Zhuge, Z. Xing, K. Zhang, T. M. Hoang, M. Morsy-Osman, M. Y. Sowailem, Y. Li, J. Wu, and D. V. Plant, “264 Gb/s twin-SSB-KK direct detection transmission enabled by MIMO processing,” in Optical Fiber communications conference (OFC) (2018), paper W4E.5.
10. H. Y. Chen, N. Kaneda, J. Lee, J. Chen, and Y. K. Chen, “Optical filter requirements in an EML-based single-sideband PAM4 intensity-modulation and direct-detection transmission system,” Opt. Express 25(6), 5852–5860 (2017). [CrossRef] [PubMed]
11. J. Lee, N. Kaneda, and Y. Chen, “112-Gbit/s intensity-modulation direct-detection vestigial-sideband PAM4 transmission over an 80-km SSMF link,” in Proceedings of European conference on optical communications (ECOC) (2016), paper M.2.D.3.
12. M. Presi, G. Cossu, G. Contestabile, E. Ciaramella, C. Antonelli, A. Mecozzi, and M. Shtaif, “Transmission in 125-km SMF with 3.9 bit/s/Hz spectral efficiency using a single-drive MZM and a direct-detection Kramers-Kronig receiver without optical CD compensation,” in Optical Fiber communications conference (OFC) (2018), paper Tu2D.3.
13. Z. Xing, D. Patel, T. Hoang, M. Qiu, R. Li, E. Fiky, M. Xiang, and D. Plant, “100Gb/s 16-QAM Transmission over 80 km SSMF using a silicon photonic modulator enabled VSB-IM/DD system,” in Optical Fiber communications conference (OFC) (2018), paper M2C.7.
14. N. Diamantopoulos, W. Kobayashi, H. Nishi, K. Takeda, T. Kakitsuka, and S. Matsuo, “56-Gb/s VSB-PAM-4 over 80-km using 1550-nm EA-DFB laser and reduced-complexity nonlinear equalization,” in Proceedings of European conference on optical communications (ECOC) (2017), paper W.4.P2.SC5.5. [CrossRef]
15. R. Bonk, “SOA for future PONs,” in Optical Fiber communications conference (OFC) (2018), paper Tu2B.4.
16. S. P. Duill, P. Landais, and L. P. Barry, “Estimation of the performance improvement of pre-amplified PAM4 systems when using multi-section semiconductor optical amplifiers,” Appl. Sci. (Basel) 7(9), 908 (2017). [CrossRef]
17. J. Zhang, J. Wey, and J. Yu, “Experimental demonstration of unequally spaced PAM-4 signal to improve receiver sensitivity to 50-gbps PON with power-dependent noise distribution,” in Optical Fiber communications conference (OFC) (2018), paper M2B.3.
18. Z. V. Rizou, K. E. Zoiros, and M. J. Connelly, “Semiconductor optical amplifier pattern effect suppression using optical notch filtering,” J. Eng. Sci. Tech. Rev. 9(4), 198–201 (2016).
19. G. P. Agrawal and N. A. Olsson, “Self-phase modulation and spectral broadening of optical pulses in semiconductor laser amplifiers,” IEEE J. Quantum Electron. 25(11), 2297–2306 (1989). [CrossRef]