Open Access
Add to BibSonomy
Abstract
In this paper, we demonstrate a self-homodyne coherent system with a significantly narrowed effective linewidth using optical carrier recovery based on stimulated Brillouin scattering (SBS), employing only coarse path length matching. The effective linewidth of the SBS-based receiver system is reduced from 75 kHz to less than 2 kHz, which is estimated by Lorentzian fitting of power spectra, and confirmed by simulation results of the tolerance window length for phase noise compensation (PNC) with different linewidth. Both experimental and numerical studies on the tracking requirements on PNC algorithms confirm effective linewidth reduction to this level, and show a 32x relaxation of the phase recovery tracking window length. This highlights the potential to significantly reduce the computational complexity of PNC even in coarsely optimized SBS-based self-homodyne coherent systems, providing an alternative to using demanding ultra-low linewidth lasers.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
With the widespread application of 5G, cloud computing, and other services, there are higher requirements for the transmission capacity in short-reach and metropolitan-area networks. The application of coherent optical communication systems to increase capacity in these networks has been widely studied [1,2] due to its higher spectral efficiency and better sensitivity levels compared to direct detection systems [3]. However, one of the potential issues in adopting coherent communications in this context is the high computational complexity of its digital signal processing (DSP). Relaxing the requirements on DSP is attractive for potential reductions in both energy usage and latency [4,5]. Phase noise and frequency offset compensation are two key stages within a coherent reception DSP flow that can be simplified by reducing the effective linewidth in a coherent system. The effective linewidth is the combined linewidth of the signal carrier and local oscillator (LO) laser, and directly relates to the magnitude of phase noise on a received signal after coherent detection.
Effective linewidth in a coherent communications system can be reduced in multiple ways. One way is to employ ultra-narrowband lasers which have reduced phase and frequency fluctuations, and can be achieved in a compact form, but require using ultra-high finesse cavities and hardware-based feedback control loops [6,7]. Alternately, phase noise and frequency offsets can be measured and compensated using compact, hardware-based feed-forward carrier recovery systems [8–10], but at the cost of requiring additional photodetectors and dynamic electro-optic modulation.
In contrast, self-homodyne coherent (SHC) detection uses the optical carrier from the signal itself as a Ref. [11–20]. This requires the implementation of optical carrier recovery (OCR), to provide an LO in which the phase noise and frequency shifts equal to that of the signal. This recovered carrier can then be used to effectively remove the phase noise and frequency offsets, by beating with the signal in a coherent receiver, as long as the phase variation between the recovered carrier and signal is correlated. In SHC systems, the difference in the path length that the signal and the recovered carrier travel over can partially decorrelate their phase variations. This can be particularly problematic for SBS-based SHC implemented in long lengths of optical fibers [18–20]. However, even a partial correlation in the phase variation between signal and recovered carrier should still result in a reduction in effective linewidth.
In this paper, we reduce the effective linewidth of an SBS-based SHC system simply by adding an extra fiber in the signal path, to coarsely match the path length of the Brillouin gain medium used for optical carrier recovery, while still leaving a path length mismatch of ∼ 150 m. Even with this path length mismatch, we observe a dramatic reduction in the phase noise of the signal. We show that this coarse path-length matching enables the computational complexity reduction of PNC, as illustrated by the ability to increase the window size of the PNC algorithm [21], over 32x from 211 to 215 points. Lorentzian fitting of power spectra and the comparison of experimental and simulation results of the tolerance to PNC window length with different linewidth both show that the effective linewidth of the system is narrowed from 75 kHz to less than 2 kHz. These results show that SBS-based OCR can be an effective tool for phase noise mitigation, even in the presence of large path length mismatches, relaxing DSP requirements on SHC systems without the need for ultra-low linewidth lasers.
2. Effective linewidth reduction in an SHC system with SBS-based OCR
The challenge in SHC systems is to enable OCR without requiring significant usable spectrum to be reserved for this operation, which compromises spectral efficiency. The SHC schemes proposed in [11–13] require one polarization to transmit the reference carrier, which results in half of the spectrum efficiency loss. Optical injection locking is another way to achieve OCR, where a guard band (often on the scale of GHz) is required due to the amplification bandwidth of injection locking, resulting in a decrease in spectral efficiency [14,15]. Stimulated Brillouin scattering (SBS) can achieve OCR with a very narrow guard band due to its tens-of-megahertz amplification bandwidth [16–20], and can be made self-tracking and transparent to wavelength [22].
For SBS-based SHC systems, a remaining issue is to match the path length travelled by the carrier in the OCR sub-system, with the path travelled by the signal, before coherent reception. This is required so that the carrier phase variation captured by OCR matches that of the signal, so that the beating between the recovered carrier and signal cancel out laser phase noise. Beating the signal with an LO with correlated phase and frequency fluctuations is what enables a lower effective linewidth in SHC systems. In order to completely remove the requirement for phase noise compensation (PNC), the product of real laser linewidth and path mismatch should be less than 0.18 MHz.m [23], to limit the optical signal-to-noise ratio (OSNR) penalty or remove PNC to under 1 dB. This means that the path mismatch length needs to be within several meters for an SHC system employing a commercial external cavity laser (ECL) with a linewidth of tens of kilohertz, to avoid phase noise compensation completely. In fiber-based SBS amplifiers, standard single-mode fibers with several kilometer-long lengths can be used as a gain medium. It is hard to precisely match the length of the signal path and the recovered carrier path in SBS-based OCR systems implemented with long optical fibers, which then typically means that such systems require traditional PNC algorithms to operate [18–20].
However, a significant reduction in the effective linewidth should still enable significant gains in lowering the computational complexity of the PNC algorithms – in a similar way to reducing DSP complexity by using ultra-narrow linewidth lasers [6,7]. Coarsely matching path lengths is feasible in an SBS-based OCR system that uses a long length of optical fiber as a gain medium. While this coarse matching might not completely eliminate the requirement for PNC, the remaining correlation between the recovered carrier and signal should enable a reduction in effective linewidth – and so enable a significant saving in computational complexity for the PNC.
The setup for an SBS-based SHC system is shown in Fig. 1. At the receiver, the incoming signal is used to regenerate the recovered optical carrier via an SBS amplifier, as shown in Fig. 1(a). A residual optical carrier is transmitted with the modulated signal and propagated through a gain medium for amplification, which is enabled by the counter-propagation of the SBS pump. The SBS pump is obtained through frequency shifting of the residual carrier, so the frequency difference between the carrier and the pump is equal to the Stokes shift for SBS amplification. The Brillouin gain medium we used in this paper is a 10 km SSMF, which will cause a significant time difference between the recovered optical carrier and the modulated signal (∼50 μs). To reduce the effective linewidth of the SHC reception by the SBS-based OCR, a second fiber spool of SSMF also labelled as being 10 km long is added in the signal path to coarsely synchronize the recovered carrier and the modulated signal (Fig. 1(b)). In this experiment, we confirm via optical time domain reflectometry that there is approximately 150 m difference in the spool lengths (representing a residual delay of ∼750 ns).
To investigate the effective linewidth reduction with the coarse synchronization, an RF pilot is modulated together with an optical carrier, and then beats in a coherent receiver with either an optical carrier recovered by the SBS-based OCR (as shown in Fig. 1(b)), or an independent ECL. The modulated RF pilot is located at a frequency of 2.8 GHz, and the beating signal is sampled by an oscilloscope running at sampling rate of 5 Gs/s with 217 sampling points, resulting in a frequency resolution of 37.3 Hz. The spectra of the resulting beating are shown in Fig. 2, showing relative power normalized to the power at the central measured beat frequency for each trace. As shown in Fig. 2(a), the effective linewidth is dramatically reduced in the SBS-based receiver, with an estimated linewidth (by Lorentzian fitting) of ∼ 2 kHz for the SBS-based receiver and ∼ 75 kHz for the ECL-based receiver. We note that without the path length matching fiber, the SBS-based receiver has a similar effective linewidth to the receiver with the independent ECL. This estimate of effective linewidth reduction is not definitive, e.g., when we zoom in to a frequency range of 50 kHz, the estimated linewidth of SBS-based receiver by Lorentzian fitting is ∼ 100 Hz, as shown in Fig. 2(b). We will further analyze the effective linewidth reduction of the SBS-based SHC receiver through comparing experimental and simulation results in Section 4.
Fig. 2. Power spectra for the beating of the RF pilot and the SBS-based recovered carrier or an independent ECL within the frequency range of (a) 50 MHz and (b) 50 kHz. LF: Lorentzian fit. Frequency axis is centered around the central measured beat frequency for each trace.
3. Experimental setup
Figure 3 shows the experimental setup for the SBS-based SHC system. An ECL with a measured linewidth of ∼37 kHz operating at 1552.52 nm is used as the optical carrier, followed by a dual-polarization (DP) IQ modulator. A Nyquist pulse-shaped modulated signal is generated by an arbitrary waveform generator (AWG) operating at 92 GSa/s, generating two sub-bands with a central guard band for SBS-based OCR. The spectrum of the modulated signal is shown in Fig. 3(a). In the experiment, DP-16QAM with a bandwidth of 20 GHz for each sub-band and DP-QPSK with a bandwidth of 23 GHz for each sub-band are employed for ML-PNC [24] and VV-PNC [25], respectively. A 1 GHz guard band is used in this experiment, which is excessive for the purpose of SBS-based OCR, and is used for diagnostic purposes only. The guard band can be reduced with a negligible penalty due to the narrow bandwidth of Brillouin amplification used for OCR [19]. The residual carrier to signal power ratio (CSPR) is adjusted by changing the bias of the modulator. The signal is transmitted through 80 km SSMF, with an Erbium-doped optical fiber amplifier (EDFA) for loss compensation.
Fig. 3. Experimental setup for the SBS-based SHC system. Inset: Optical spectrum of (a) the modulated signal with residual optical carrier; (b) input signal of the SBS-based OCR; and (c) recovered optical carrier by the SBS-based OCR.
At the receiver side, the signal is first split by a 99:1 coupler, with 99% being sent to the signal port of the coherent receiver, and 1% sent to the SBS-based OCR sub-system. A 50:50 coupler within the SBS-based OCR sub-system splits the signal into two branches, where in the upper branch the residual optical carrier is forward propagated in the Brillouin gain medium to undergo selective amplification, while in the lower branch the residual carrier is amplified and then frequency shifted by an IQ modulator at 10.82 GHz to be used as the SBS pump, backward propagating in the SBS gain medium. The SBS pump is amplified to 26.5 dBm by a second EDFA to achieve an SBS gain of 21 dB in the 10 km SSMF. Carrier and SBS pump are co-polarized via polarization controllers (PCs) to maximize SBS gain, where the optical spectra are shown in Figs. 3(b) and 3(c), respectively. Note that there is a second peak at the frequency gap of 10.82 GHz as shown in Fig. 3(c), which is the Rayleigh backscatter of the SBS pump and is 34 dB lower than the recovered carrier.
After the OCR stage, the recovered optical carrier is fed into the coherent receiver used as an LO. As earlier, we roughly synchronize the signal with the recovered carrier via a spool of SSMF, leaving ∼150 m path length mismatch. An 80 GSa/s oscilloscope samples the outputs of the coherent detector, and the signal is processed offline. The signal is first down-shifted by a set center frequency value for each sub-band and then resampled to 2 Sa/symbol, followed by chromatic dispersion compensation. A 2.5% roll-off root-raided cosine (RRC) filter which is the same as that in the transmitter-side DSP is used for matched filtering. Then for DP-16QAM signal, frame synchronization is implemented, followed by channel equalization, ML-PNC, and bit error ratio (BER) counting. For DP-QPSK signal, frame synchronization is implemented for BER counting after blind channel equalization and VV-PNC. In all presented results, we show the averaged performance (Q2, BER) of the two polarization channels of the signals. We compare this to when an independent ECL is used as LO as a benchmark for performance. The same DSP flows are used for the ECL-based receiver, except that frequency offset compensation is employed before matched filtering. A key parameter to optimize in a self-coherent system is the carrier-to-signal power ratio (CSPR). A residual optical carrier with low CSPR will cause insufficient SNR of the SBS-based OCR, resulting in a decrease in system performance. While with high CSPR, the OSNR of the modulated signal will decrease after transmission. We search for the optimal CSPR value for SBS-based receiver by adjusting the bias of the DP-IQ modulator in back-to-back setup, as shown in Fig. 4. Hence, we set the CSPR at the optimal value around -8.5 dB, which we use throughout the rest of this investigation. This low value of CSPR can be achieved using only a small percentage of Vπ as bias (∼2%), and so provides a very small reduction of the linear modulation region.
4. Results
4.1 Phase noise tolerance with ML-PNC
To investigate the impact of effective linewidth reduction with the SBS-based SHC receiver, we first study the tolerance of the ML-PNC window length with and without SBS-based OCR. An increase in the window length of a given PNC algorithm means that the phase noise is estimated by averaging over a longer period of time. A larger window means that the PNC algorithm will only effectively track slower phase fluctuations, and so will only work well in systems with lower effective linewidths. More broadly, successful operation with longer window lengths means that a PNC with less computational complexity can be used.
Figure 5(a) shows the curves of Q2 factor versus the ML-PNC window length for DP-16QAM signal in back-to-back setup. Without the SBS-based OCR, due to the randomness of phase noise, the Q2 factor of each captured data-stream can vary greatly as the ML-PNC becomes ineffective (i.e., with larger estimation windows). As such, we represent the Q2 factor of each captured data-stream by triangle points with same color, with the average of all captured data-streams plotted by the dashed line. It is obvious that for ECL-based receivers, when the window length is greater than 211 (2048) points, the Q2 factor begins to decrease. For SBS-based SHC receiver, the Q2 factor varies by less than 0.5 dB as the window length increases – we are limited to a PNC window of 215 points due to the memory length of our transmitter-side AWG. This indicates that the SBS-based SHC receiver system can greatly reduce the effective linewidth by roughly synchronizing the signal with the recovered carrier, so that the phase noise changes slowly. This is essential to reduce the computational complexity of the phase noise compensation module, especially in real-time systems [21]. Figure 5(b) shows simulation results, with different effective linewidths applied at a set OSNR of 23.5 dB. In the simulation, the double-band DP-16QAM with a bandwidth of 20 GHz for each sub-band is first up-sampled to 92 GSa/s to simulate the sampling rate of the AWG, and then the channel impairments, such as the OSNR and linewidth, are added. We set the OSNR as 23.5 dB so that the simulation results are as close as possible to the experimental results as a reference. Then, the signal is first down-sampled to 80 Gsa/s, which is the sampling rate of the oscilloscope, followed by the same DSP used in the experiment. The highest value of effective linewidth that shows the same negligible degradation as the SBS-based SHC system was 2 kHz. However, as we could not observe an obvious performance degradation even when using the maximum window length possible in our experiment, we can only infer that the effective linewidth of the SBS-based SHC system is better than 2 kHz. We note that the independent ECL result does not precisely correspond to a specific simulated linewidth, this might be related to the limited number of runs available from the experiment.
Fig. 5. (a) Experimental results of Q2 factor versus window length for 16QAM signal employing ML-PNC. (b) Simulation results of Q2 factor versus window length for 16QAM signal employing ML-PNC with different linewidth.
Next, we compare the performance for DP-16QAM using the SBS-based receiver and the ECL-based receiver with noise loading in back-to-back setup. As before, we present the Q2 factor results from different capture when using an ECL-based receiver and PNC of 215 points window length with triangle points in Fig. 6(a), and the same as in the following figures. The Q2 factor penalty at each OSNR value is within 0.5 dB when using SBS-based receiver compared to ECL-based receiver with an ML-PNC window length of 27 points, as shown in Fig. 6(a). However, while we enlarge the window length of ML-PNC to 215 points, the impact of different effective system linewidths is highlighted. The Q2 factor penalty is less than 0.5 dB for SBS-based receiver, while it is 6.2 dB for ECL-based receiver with OSNR of 31 dB, which indicates that the phase changes are too large over the 215-point window length to be effectively tracked by the ML-PNC. Figure 6(b) shows the BER performance comparison, where the OSNR penalty is ∼1.2 dB and ∼0.4 dB for the SBS-based receiver compared with the ECL-based receiver for 7% FEC threshold and 20% FEC threshold, respectively. After broadening the ML-PNC window length to 215 points, there is only 0.2 dB OSNR penalty compared with window length of 27 points for 20% FEC threshold, and 2.1 dB OSNR penalty for 7% FEC threshold for the SBS-based receiver. For the ECL-based receiver, it is hard to reach the 20% FEC threshold with such a large PNC window length.
Fig. 6. Performance comparison for 16QAM signal employing ML-PNC in back-to-back measurements. (a) Q2 factor versus OSNR. (c) BER versus OSNR.
We also evaluate performance after 80 km SSMF transmission for DP-16QAM, to ascertain that the SBS-based system can operate well in fiber links, with results shown in Fig. 7. The Q2 factor difference when using 27 and 215 ML-PNC windows with the SBS-based receiver is less than 0.5 dB. In Fig. 7(b), the required OSNR is 24.6 dB and 17.2 dB for 7% and 20% indicative FEC thresholds with 215 points window length of ML-PNC, and only 0.8 dB OSNR penalty is induced in both cases compared with the back-to-back setup, which indicates that the impact of transmission on the tolerance to the window length of ML-PNC is negligible. Figures 8(c) and 8(d) show the signal constellation of one set of data detected by ECL-based receiver with 27 and 215 points window length and 30.2 dB OSNR, which shows there is residual phase noise that cannot be compensated with a large window length. While the signal constellation for SBS-based receiver with the same setting implies that the change of phase noise is still within the capability of ML-PNC with 215 points window length, as shown in Figs. 7(e) and 7(f).
Fig. 7. Performance comparison for 16QAM signal employing ML-PNC after 80 km SSMF transmission. (a) Q2 factor versus OSNR. (b) BER versus OSNR. Signal constellation for ECL-based LO receiver with PNC window length of (c) 27 points and (d) 215 points, and for SBS-based OCR receiver with PNC window length of (e) 27 points and (f) 215 points.
Fig. 8. (a) Experimental results of Q2 factor versus window length for QPSK signal employing VV-PNC. (b) Simulation results of Q2 factor versus window length for QPSK signal employing VV-PNC with different linewidth.
4.2 Phase noise tolerance with VV-PNC
Since the window length of ML-PNC is limited by the memory length of the AWG in the transmitter-side, we adopt VV-PNC algorithm for DP-QPSK format to further broaden the PNC window length to investigate the effective linewidth of the system. Since the channel equalization and VV-PNC are both based on blind estimation algorithms, we do not need to retrieve a frame of data from the received continuous signal, so that the window length of PNC is no longer limited to the frame length of the transmitted data. As shown in Fig. 8(a), we further increase the length of the PNC window to 217 (131072) points to show the system performance with different effective linewidth. For the SBS-based SHC system, the Q2 factor penalty is within 0.5 dB. However, for the ECL-based traditional coherent system, the VV-PNC maximum window length that can be stably demodulated is 213 points. When the VV-PNC window continues to increase, the signal recovery performance becomes unstable and the system performance begins to decline due to the randomness of the phase noise. Figure 8(b) shows the simulation results under different linewidths. When the window length is increased to 217 points, the Q2 factor of the system has a < 0.5 dB penalty with a linewidth of 200 Hz, a 1.7 dB penalty with a linewidth of 500 Hz, and a 4 dB penalty with a linewidth of 2 kHz. This suggests that the effective linewidth of the SBS-based SHC system is less than 2 kHz, and maybe as low as 200 Hz – closer to the ‘zoomed in’ fitting of Fig. 2(b) than the estimate from fitting in Fig. 2(a).
For QPSK format, the Q2 factor performance shown in Fig. 9(a) is similar to that of 16QAM shown in Fig. 6(a), which shows that the effective linewidth of the SBS-based system is within the tolerance of VV-PNC with a window length of 215 points. The required OSNR is 9.9 dB and 10.1 dB for ECL-based receiver and SBS-based receiver respectively with a 7% FEC threshold, as shown in Fig. 9(b), and 0.1dB OSNR penalty is induced by enlarging the window length of VV-PNC from 27 points to 215 points in SBS-based receiver system. Similar to 16QAM signal, the QPSK signal with VV-PNC window length of 215 points in ECL-based receiver system cannot be successfully recovered even with 20% FEC and large OSNR.
Fig. 9. Performance comparison for QPSK signal employing VV-PNC in back-to-back measurements. (a) Q2 versus OSNR. (b) BER versus OSNR.
Figure 10 shows the performance evaluation for DP-QPSK signals after 80 km SSMF transmission. For both the ECL- and SBS-based receiver, 0.2 dB Q2 penalty is observed at OSNR of 30 dB with VV-PNC window length of 27 points compared to the back-to-back measurements. The corresponding constellation for ECL-based and SBS-based receiver are shown in Figs. 10(c) and 10(e), respectively. Additional 0.42 dB Q2 penalty is induced by enlarging the VV-PNC window length to 215 points for SBS-based receiver, and the constellation is shown in Fig. 7(e). The required OSNR is 11.2 dB for both ECL-based receiver and SBS-based receiver with a 7% FEC threshold, where 1.1 dB Q2 penalty is induced by the transmission. However, enlarging the window length in SBS-based receiver system does not bring additional Q2 penalty. Similar to back-to-back measurement, signal detected by ECL-based receiver cannot be successfully recovered with VV-PNC window length of 215 points. Figure 10(d) shows the constellation of one set of data with OSNR of ∼ 31 dB, detected by ECL-based receiver and demodulation with VV-PNC window length of 215 points. Although the calculated average of Q2 factor is 8.6 dB in this case, the BER is still larger than 2e-2 due to the residual phase noise that cannot be compensated by VV-PNC with such a large window length.
Fig. 10. Performance comparison for QPSK signal employing VV-PNC after 80 km SSMF transmission. (a) Q2 factor versus OSNR. (b) BER versus OSNR. Signal constellation for ECL-based LO receiver with PNC window length of (c) 27 points and (d) 215 points, and for SBS-based OCR receiver with PNC window length of (e) 27 points and (f) 215 points.
5. Discussion
According to the Lorentzian fitting of power spectra and the comparison of experimental and simulation results on the impact of linewidth on tolerance of PNC window length, we show the effective linewidth of the SBS-based system is reduced from 75 kHz to less than 2 kHz. Although the product of laser linewidth of 37 kHz and path mismatch of 150 m is 5.55 MHz·m which is much larger than 0.18 MHz·m stated in [23], we still see a significant reduction in the requirements on the PNC DSP with both maximum-likelihood algorithm and blind algorithm. We attribute this simply to the reduced effective linewidth. We do note that the shape of the beat spectrum does not match a pure Lorentzian shape, and so are cautious in our interpretation of what precisely the effective linewidth of this system really is. However, the match between the simulated systems results and experiment suggest something in the range of 200Hz-2kHz is likely. Reducing the effective linewidth below this level is likely possible when using a more compact Brillouin gain platform than optical fiber, which would suggest that it would be advantageous to investigate effective linewidth reduction in systems that use photonic chips for the Brillouin amplification step (e.g. [16,17]). We also note that this value of reduced linewidth is only demonstrated for our ECLs, with a measured linewidth of ∼37 kHz. How this changes with different input laser linewidth requires further investigation.
There are a large number of phase noise compensation algorithms either in use or proposed for use in optical communications systems. In this investigation, we have compared performance with both a Viterbi-Viterbi algorithm (VV-PNC), and a training-based maximum likelihood algorithm (ML-PNC), which uses precise knowledge of the sent data pattern to track phase. The ML-PNC represents the best-case scenario for phase compensation, while the VV-PNC is simple in implementation. We would expect our results are generalizable to other PNC algorithms, which we would expect to fall somewhere between VV-PNC and ML-PNC in terms of tolerance to residual effective linewidth after self-homodyne detection. While we have established that this compensation scheme with significant mismatch in the carrier recovery and signal paths can relax requirements on the PNC algorithms shown, the exact impact on different algorithms requires further investigation. Moreover, there is clear scope to investigating purpose-designed low computational complexity algorithms taking advantage of the SHC approach.
While the SBS-based SHC approach we investigate here is useful in reducing effective linewidth, and does not require ultra-low linewidth lasers, there are several practical challenges to be overcome before implementation in a commercial system. Polarization alignment of the incoming signal to the modulators operational axis is a key requirement to operation, as well as subsequent alignment to the polarization axis of the coherent receiver. One possible solution is the addition of polarization agnostic techniques, such as those presented in [14], that might help to solve this issue. Solving this problem would represent a significant step toward the practical implementation of an SBS-based SHC scheme, as we have shown here that the requirements on path length matching can be significantly relaxed, and we have recently shown that the wavelength dependence of SBS gain can be overcome to enable wavelength agnostic operation [22].
6. Conclusion
In this paper, we show the effective linewidth reduction of an SBS-based SHC system from 75 kHz to less than 2 kHz by only coarse path length matching. Experimental results show that the tolerance of the ML-PNC and VV-PNC window length can be increased up to 215 points (limited by the memory length of AWG) and 217 points, respectively. System simulation results also indicate that the system effective linewidth should be less than 2 kHz. A relaxation of phase noise compensation algorithm can be achieved with such a narrow effective linewidth, with significant computational complexity reduction of PNC in DSP.
Funding
Australian Research Council (DP190100992); Air Force Office of Scientific Research (FA2386-16-1-4036); Office of Naval Research Global (N62909-18-1-2013); Science and Engineering Research Board (SRG/2019/001632).
Disclosures
The authors declare that there are no conflicts of interest related to this article.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. J. K. Perin, A. Shastri, and J. M. Kahn, “Coherent Data Center Links,” J. Lightwave Technol. 39(3), 730–741 (2021). [CrossRef]
2. M. Morsy-Osman and D. V. Plant, “A Comparative Study of Technology Options for Next Generation Intra- and Inter-datacenter Interconnects,” Proc. Optical Fiber Communication Conference, W4E.1 (2018).
3. K. Kikuchi, “Fundamentals of coherent optical fiber communications,” J. Lightwave Technol. 34(1), 157–179 (2016). [CrossRef]
4. J. K. Perin, A. Shastri, and J. M. Kahn, “Design of Low-Power DSP-Free Coherent Receivers for Data Center Links,” J. Lightwave Technol. 35(21), 4650–4662 (2017). [CrossRef]
5. M. Morsy-Osman, M. Sowailem, E. El-Fiky, T. Goodwill, T. Hoang, S. Lessard, and D. V. Plant, “DSP-free ‘coherent-lite’ transceiver for next generation single wavelength optical intra-datacenter interconnects,” Opt. Express 26(7), 8890–8903 (2018). [CrossRef]
6. W. Jin, Q.-F. Yang, L. Chang, B. Shen, H. Wang, M. A. Leal, L. Wu, M. Gao, A. Feshali, M. Paniccia, K. J. Vahala, and J.E. Bowers, “Hertz-linewidth semiconductor lasers using CMOS-ready ultra-high-Q microresonators,” Nat. Photonics (2021).
7. G. M. Brodnik, M. W. Harrington, D. Bose, A. M. Netherton, W. Zhang, L. Stern, P. A. Morton, J. E. Bowers, S. B. Papp, and D. J. Blumenthal, “Chip-Scale, Optical-Frequency-Stabilized PLL for DSP-Free, Low-Power Coherent QAM in the DCI,” Proc. Optical Fiber Communication Conference, M3A.6 (2020).
8. S. Beppu, K. Kasai, M. Yoshida, and M. Nakazawa, “2015 QAM (66 Gbit/s) single-carrier coherent optical transmission over 150 km with a potential SE of 15.3 bit/s/Hz,” Opt. Express 23(4), 4960–4969 (2015). [CrossRef]
9. Z. Xuan and F. Aflatouni, “Integrated coherent optical receiver with feed-forward carrier recovery,” Opt. Express 28(11), 16073–16088 (2020). [CrossRef]
10. R. T. Watts, S. G. Murdoch, and L. P. Barry, “Phase noise reduction of an optical frequency comb using a feed-forward heterodyne detection scheme,” IEEE Photonics J. 8(1), 1–7 (2016). [CrossRef]
11. M. Y. S. Sowailem, E. El-Fiky, M. Morsy-Osman, Q. Zhuge, T. M. Hoang, S. Paquet, C. Paquet, I. Woods, O. Liboiron-Ladouceur, and D. V. Plant, “Self-homodyne system for next generation intra-datacenter optical interconnects,” Opt. Express 25(22), 27834–27844 (2017). [CrossRef]
12. R. Kamran, S. Naaz, S. Goyal, and S. Gupta, “High-Capacity Coherent DCIs Using Pol-Muxed Carrier and LO-Less Receiver,” J. Lightwave Technol. 38(13), 3461–3468 (2020). [CrossRef]
13. M. Nakamura, Y. Kamio, and T. Miyazaki, “Real-time 40-Gbit/s 16-QAM self-homodyne using a polarization- multiplexed pilot-carrier,” Proc. IEEE/LEOS Summer Topical Meetings, WC2.3 (2008).
14. J. Jignesh, B. Corcoran, J. Schröder, and A. Lowery, “Polarization independent optical injection locking for carrier recovery in optical communication systems,” Opt. Express 25(18), 21216–21228 (2017). [CrossRef]
15. J. Jignesh, A. Lowery, and B. Corcoran, “Inter-channel nonlinear phase noise compensation using optical injection locking,” Opt. Express 26(5), 5733–5746 (2018). [CrossRef]
16. E. Giacoumidis, A. Choudhary, E. Magi, D. Marpaung, K. Vu, P. Ma, D.-Y. Choi, S. Madden, B. Corcoran, M. Pelusi, and B. J. Eggleton, “Chip-based Brillouin processing for carrier recovery in self-coherent optical communications,” Optica 5(10), 1191–1199 (2018). [CrossRef]
17. A. Choudhary, M. Pelusi, D. Marpaung, T. Inoue, K. Vu, P. Ma, D.-Y. Choi, S. Madden, S. Namiki, and B. J. Eggleton, “On-chip Brillouin purification for frequency comb-based coherent optical communications,” Opt. Lett. 42(24), 5074–5077 (2017). [CrossRef]
18. M. Pelusi, T. Inoue, and S. Namiki, “Enhanced Carrier to Noise Ratio by Brillouin Amplification for Optical Communications,” J. Lightwave Technol. 38(2), 319–331 (2020). [CrossRef]
19. E. Giacoumidis, E. Magi, A. Choudhary, D. Marpaung, B. Corcoran, M. Pelusi, and B. J. Eggleton, “Enhanced Self-Coherent Optical OFDM using Stimulated Brillouin Scattering,” Proc. Optical Fiber Communication Conference, Th4I.2 (2017).
20. A. Lorences-Riesgo, M. Mazur, T. A. Eriksson, P. A. Andrekson, and M. Karlsson, “Self-homodyne 24×32-QAM superchannel receiver enabled by all-optical comb regeneration using brillouin amplification,” Opt. Express 24(26), 29714–29723 (2016). [CrossRef]
21. S. M. Bilal, C. R. S. Fludger, V. Curri, and G. Bosco, “Multistage Carrier Phase Estimation Algorithms for Phase Noise Mitigation in 64-Quadrature Amplitude Modulation Optical Systems,” J. Lightwave Technol. 32(17), 2973–2980 (2014). [CrossRef]
22. A. Zarifi, M. Merklein, Y. Liu, A. Choudhary, B. J. Eggleton, and B. Corcoran, “Wide-range optical carrier recovery via broadened Brillouin filters,” Opt. Lett. 46(2), 166–169 (2021). [CrossRef]
23. X. Zhou, Y. Gao, J. Huo, and W. Shieh, “Theoretical Analysis of Phase Noise Induced by Laser Linewidth and Mismatch Length in Self-Homodyne Coherent Systems,” J. Lightwave Technol. 39(5), 1312–1321 (2021). [CrossRef]
24. X. Zhou, “An Improved Feed-Forward Carrier Recovery Algorithm for Coherent Receivers With M-QAM Modulation Format,” IEEE Photonics Technol. Lett. 22(14), 1051–1053 (2010). [CrossRef]
25. D.-S. Ly-Gagnon, S. Tsukamoto, K. Katoh, and K. Kikuchi, “Coherent Detection of Optical Quadrature Phase-Shift Keying Signals With Carrier Phase Estimation,” J. Lightwave Technol. 24(1), 12–21 (2006). [CrossRef]
