Open Access
Add to BibSonomy
Abstract
We propose the use of Manchester encoding in conjunction with balanced detection to overcome the mode partition noise (MPN) limit of quantum-dash Fabry-Perot mode-locked lasers (QD-MLLs) used as multi-wavelength sources in short-reach applications. The proposed approach is demonstrated for a 10-mode laser, each carrying a 10-Gb/s signal. We show that bit-error-rate floors as high as 10−4 when traditional non-return-to-zero (NRZ) modulation is employed with a single-ended detection scheme can be pushed below 10−9 thanks to the introduction of Manchester encoding together with balanced detection. The benefit of the scheme could be attributed to the spectral shift of the Manchester spectrum, resulting in a smaller overlap with the high-relative intensity noise (RIN) region present at low frequencies, and the use of balanced detection. We clarify the origin of the performance improvement through comparisons of single-ended and balanced detection and the use of a RIN emulation technique. We unambiguously show that the use of balanced detection plays the leading role in MPN mitigation enabled by Manchester modulation.
© 2017 Optical Society of America
1. Introduction
Quantum-dot or quantum-dash (QD) mode-locked laser (MLL) sources are highly attractive in the context of short-reach access and data-centre networks since they enable the generation of an entire comb of wavelength division multiplexed (WDM) channels with equal frequency spacing in a single device [1,2]. Those compact sources, which have been used in a number of transmission experiments [3–6], are envisaged as a replacement for full arrays of single-mode lasers. They also have the potential to be heterogeneously integrated on the silicon platform through wafer bonding techniques [7] in order to implement a fully integrated WDM transmitter. In this case, a transmitter architecture where the QD-MLL is followed by a number of silicon ring-resonator modulators (RRMs) in series, each modulating a given line of the generated frequency comb, is particularly attractive since it does not require any wavelength multiplexer and demultiplexer at the transmitter side [8–11].
One well-known limitation of such sources, however, is the increase of the relative intensity noise (RIN) of the filtered comb lines due to the effect of mode partition noise (MPN) [12]. Due to the competition for a common injected carrier population between the different modes, the RIN of individually filtered laser modes is larger than the RIN resulting from the simultaneous detection of all modes. This excess RIN present on each mode results in potentially poor bit-error-rate (BER) performance, including the occurrence of BER floors, when the QD-MLL is used as part of a multi-channel transmitter where the different lines are individually modulated by external modulators. While this effect may not be significant in systems employing forward-error correction (FEC) with target BER of the order of 10−3 or higher, it is certainly detrimental when the target BER is 10−9 or lower, as is the case in some short-reach applications where the excess power consumption and latency of FEC circuits may not be acceptable at the present time. Several techniques have been proposed to mitigate the effect of MPN [13–15]. One powerful technique exploits the high-pass behaviour of a saturated semiconductor optical amplifier (SOA), which filters the low-frequency portion of the RIN spectrum, where the noise is enhanced as a result of MPN [16,17]. However, this method can only be applied on a single continuous-wave (CW) laser line, i.e. after the comb lines have been wavelength-demultiplexed and before the modulation is applied. It is therefore not compatible with the aforementioned serial modulator transmitter architecture whose main benefit is not to require demultiplexing prior to modulation.
While the QD-MLL followed by RRMs in series transmitter structure is currently the object of several implementation efforts, the potentially detrimental effect of MPN and the inherent difficulty to mitigate it in this architecture have not been considered so far. We have recently demonstrated how this issue could be circumvented by employing Manchester encoding in conjunction with balanced detection [18], although at the expense of a doubled required modulation bandwidth. We had speculated that the measured resilience to MPN could have two origins. First, the use of balanced detection is well know to improve the tolerance towards intensity fluctuations, including RIN [19]. However, balanced-detection is not compatible with the standard non-return-to-zero (NRZ) on-off keying (OOK) modulation format. In contrast, the use of Manchester encoding is fully compatible with balanced detection [20] and offers the added benefit that the spectral content of the signal is shifted towards higher frequencies, thus reducing its overlap with the spectral region where the RIN of individually filtered lines is enhanced as a result of MPN. This latter argument has in the meantime been exploited to interpret the better tolerance of Manchester and 8B/10B-encoded 4-level pulse amplitude modulation (PAM4) signals generated from a QD-MLL and detected with a single-ended receiver [21]. In this paper, we extend our results on the use of Manchester encoding in conjunction with balanced detection for OOK signals generated from a filtered QD-MLL frequency comb. In particular, we use an emulated RIN source to clarify the origin of the observed performance improvement and show that the use of Manchester encoding alone, i.e. alongside standard single-ended detection, is unable to mitigate the effect of MPN, i.e. push the RIN-induced BER floor below the target BER value of 10−9.
The rest of this paper is organized as follows. First, the MPN mitigation technique is presented in more detail, followed by a description of the experimental set-up used to compare NRZ and Manchester encoding with single-ended and balanced detection. It is shown experimentally using a real-time direct detection receiver that does not employ any form of digital signal processing (DSP) that the sensitivity to MPN of a 10-channel QD-MLL source is significantly reduced in a 10-Gb/s experiment, resulting in improved receiver sensitivity or the suppression of MPN-induced bit-error-rate floors down to 10−10 when Manchester encoding and balanced detection are simultaneously used. Then, the emulated RIN source is described and it is shown how it can be used to flexibly adapt the RIN spectrum of a continuous-wave laser source in order to study its impact on the performance of NRZ and Manchester-encoded signals with either single-ended or balanced detection.
2. Principle of the technique
A WDM transmitter consisting of a QD-MLL laser followed by a serial array of micro-ring resonator (MRR) modulators [8–10] is schematically represented in Fig. 1. Such a structure is practically attractive since it does not require demultiplexing of the comb lines at the transmitter in order to perform modulation. It also has the potential of being heterogeneously integrated on the silicon platform. Its operating principle consists in individually detuning the resonance of each MRR upon the application of a modulating signal so that it modifies the intensity of a single line of the frequency comb generated in the QD-MLL. However, since the individual lines are not accessible separately prior to modulation, the technique consisting in using a saturated SOA in order to mitigate the MPN effect cannot be applied in this transmitter architecture [16,17]. The different channels are only separated prior to detection at the receiving side of the link, where the use of a saturated SOA would also suppress the intensity modulation. New MPN mitigation techniques that are compatible with the transmitter architecture of Fig. 1 therefore need to be found.
The use of balanced detection is known to be able to reduce the impact of RIN in analogue links [19]. However, such a detection scheme is not compatible with the simple NRZ intensity modulation format widely used in cost-effective short reach digital systems. On the other hand, Manchester line coding, in which the binary data is encoded in low-to-high or high-to-low transitions within the bit slot is fully compatible with intensity modulation and balanced direct detection. This, however, results in a doubled modulation bandwidth compared to NRZ at the same bit rate and a shift of the spectral content of the modulated signal away from the low frequency region, as can be seen in Fig. 2, where the measured radio-frequency (RF) spectra of conventional NRZ and Manchester-encoded signals at 10 Gb/s are compared.
Since MPN results in an enhancement of the RIN that is particularly pronounced in the low-frequency part of the spectrum [16], it has been suggested that, by reducing the overlap between the RIN spectrum and the modulation spectral content, the use of Manchester encoding could result in a better resilience to MPN [18]. The same effect has been exploited in the context of mitigation of Rayleigh noise in bidirectional links [22], as well as for orthogonal modulation labelling [23]. In the following, we demonstrate the benefit of Manchester encoding with balanced detection for MPN mitigation in a proof-of-concept experiment. In the absence of an integrated source including a QD-MLL and MRR modulators in series, we use a stand-alone QD-MLL [1, 17] and a standard lithium-niobate Mach-Zehnder modulator to implement the concept and assess the relative performance of NRZ and Manchester encoding with respect to MPN.
3. Experimental set-up
The experimental set-up is represented in Fig. 3. A QD-MLL chip [17] biased with a current of 150 mA, generates the frequency comb with 100-GHz channel spacing shown in Fig. 4(a). In the following, the ten consecutive comb lines presenting the highest signal-to-noise ratios that are labelled in Fig. 4(a) will be exploited as a WDM source. The RIN of the source is first characterized by measuring the radio-frequency spectrum of the signal at the output of the laser after photodetection using an electrical spectrum analyzer and correcting for the effect of shot-and thermal-noise [24]. The same set-up is used for characterization of the entire comb and of selected modes after optical bandpass filtering with a 0.2-nm flat-top filter, without resorting to optical amplification. The impact of MPN is illustrated in Fig. 4(b), where the RIN spectra measured following direct detection of all the laser modes, as well as after individual detection of a single mode after bandpass filtering (worst case mode 8 & good case mode 9) are represented. The RIN is also compared to that of a reference single-mode external cavity laser (ECL) used later in the experiment. It is clear that filtering the modes results in a significant increase of the RIN, especially in the low-frequency part of the spectrum. The frequency comb is first amplified by an erbium-doped fibre amplifier (EDFA). The comb lines are then individually filtered one at a time by an optical bandpass filter (OBPF) before being intensity-modulated at 10 Gb/s in a Mach-Zehnder modulator (MZM) using either the NRZ or Manchester line code. The signal is finally detected in a preamplified receiver before BER measurement. For NRZ coding, a standard 10-Gb/s receiver is used, while a balanced detection scheme [20] making use of a pair of 30-GHz photodiodes followed by a clock and data recovery (CDR) module and a D-flip-flop is employed in the Manchester case. The operation of the balanced receiver is as follows. The Manchester signal is first duplicated in a 3-dB coupler. One of the signal copies is directly detected in one of the balanced photodiodes while the other one is delayed by half a symbol duration before being detected in the other photodiode. The signal information is therefore converted from the transitions of the Manchester signal (high-to-low or low-to-high, respectively) to a negative or positive voltage, respectively, during the second half of the symbol duration after balanced detection. The receiver therefore acts as a transition-to-binary level converter after sampling at the symbol rate in the second half of each symbol. A pseudo-random binary sequence (PRBS) length of 231 − 1 is used throughout this work. It is important to point out that all the BER measurements involving the QD-MLL source have been carried out without modifying the optical coupling condition nor the bias current in the course of the experiment. This procedure was adopted as it was observed that the BER performance of any filtered mode could vary considerably by merely changing the optical coupling. All modes have therefore been characterized under the same coupling and bias current conditions, as would be the case when employing a packaged device.
Fig. 3 Experimental set-up. QD-MLL: quantum-dash mode-locked laser; EDFA: erbium-doped fiber amplifier; OBPF: optical bandpass filter; MZM: Mach-Zehnder modulator; T/2: half-symbol delay; CDR: clock and data recovery; DFF: D flip-flop.
Fig. 4 (a) QD-MLL optical spectrum for a bias current of 150 mA. (b) Measured RIN for the joint detection of all QD-MLL modes, after filtering of modes 8 or 9, and for the reference case of an external cavity laser.
4. Results and discussion
The eye diagram of the NRZ signal obtained when an ECL is substituted to the QD-MLL source as well as of a filtered comb line (mode 8) are shown in Figs. 5(a) and 5(b) respectively. The impact of MPN is clearly visible in the latter case. The corresponding eye diagrams obtained with Manchester coding are also represented in Figs. 5(c) and 5(d). Note that both NRZ and Manchester eye diagrams are monitored with the same single photodiode.
Fig. 5 Eye diagrams of the NRZ signal obtained from (a) a reference ECL or (b) the filtered QD-MLL. (c) and (d): corresponding eye diagrams with Manchester coding.
Curves representing the measured BER as a function of the optical pre-amplifier input power for the filtered modes as well as for the reference ECL are represented in Figs. 6(a) and 6(b) for NRZ and Manchester coding, respectively. A large dispersion of the BER performance can be observed in the NRZ single-ended detection case, with the occurrence of BER floors at values as high as 10−4 for the worst-case mode 8. The best-case mode 2 exhibits more than 2.4 dB power penalty at a BER of 10−9 with respect to the ECL reference case. In contrast, the BER performance for all the filtered modes is significantly improved when Manchester coding associated to a balanced detection scheme is employed. The worst-case mode 7 shows 2.2 dB power penalty at a BER of 10−9 and its BER floor appears below 10−10.
Fig. 6 BER versus average received power for (a) NRZ and (b) Manchester encoding. The average power is measured at the input of the optical pre-amplifier.
In order to clarify the origin of the performance improvement, the measured BERs of Manchester line coding are compared in Fig. 7 for both balanced (BD) and single-ended detection (SD; in a 30-GHz photodiode) to that of NRZ line coding in the ECL reference case as well as for the filtered mode 8. The benefit of balanced detection is clearly visible in the reference case since it provides a 2.5 dB better sensitivity (at a BER of 10−9) for Manchester compared to NRZ, while Manchester encoding suffers from a 2.6 dB power penalty in the single-ended detection case. In the presence of MPN (filtered mode 8), the introduction of Manchester encoding with single-ended detection is not sufficient to suppress the BER floor below the target BER value of 10−9. It therefore appears that the use of balanced detection is largely responsible for the performance improvement of Manchester with respect to NRZ. In order to gain a better understanding of how Manchester encoding and balanced detection contribute to this improvement, as well as to scale the performance improvement with the RIN values, an emulated RIN source allowing the synthesis of RIN spectra with different shapes and levels will be used.
Fig. 7 Comparison of the BER performance for single-ended and balanced detection Manchester coding, as well as NRZ coding in the reference ECL case and for a filtered QD-MLL mode.
5. RIN emulation source
In order to emulate realistic RIN profiles, the experimental set-up depicted in Fig. 8 is used. It consists of an externally modulated ECL and an optical white noise source. An arbitrary waveform generator (AWG) acting as digital-to-analog converter is fed with amplitude-shaped white noise to mimic the low-frequency part of the RIN spectrum. The amplitude response is extracted from an experimentally measured RIN spectrum. The electrical signals are generated at a sampling rate of 12 GS/s and occupy the frequency band DC-6 GHz, which is sufficient for the measured RIN spectra to converge to their high-frequency constant levels. A Mach-Zehnder modulator is then used to imprint the electrical signals on the input ECL signal through intensity modulation so that the low frequency part of the RIN spectrum is obtained. The high-frequency constant level part of the RIN spectrum is obtained thanks to a white noise source relying on amplified spontaneous emission (ASE) noise generated by an EDFA. An optical coupler is then used to combine the two emulated parts of the RIN spectrum. Figure 9(a) illustrates how the RIN spectrum of a QD-MLL filtered mode is emulated using the proposed mechanism. It shows the measured RIN spectra at the output of the Mach-Zehnder modulator, at the output of the white noise source, as well as at the output of the whole emulator. The emulated RIN is furthermore compared to that measured on the target mode of the QD-MLL.
Fig. 8 Experimental set-up used to emulate the RIN. The low frequency part of the RIN spectrum synthesized in an arbitrary waveform generator (AWG) is modulated on a low-RIN continuous-wave (CW) optical carrier using a Mach-Zehnder modulator (MZM) and added to white noise generated in an amplified spontaneous emission (ASE) source.
Fig. 9 (a) Measured RIN spectra at the output of the externally modulated laser (corresponding to output 1 in Fig. 8), at the output of the ASE source (output 2) and at the output of the whole RIN emulating system (output 3), superimposed with the RIN spectrum of mode 7 of the QD-MLL source. (b) Comparison of BER performance of two QD-MLL filtered modes and their corresponding RIN emulated sources.
To ensure that a genuine QD-MLL filtered mode and an emulated RIN source sharing the same RIN spectrum provide similar performance, Fig. 9(b) compares the BER curves obtained from two 10-Gb/s NRZ modulated QD-MLL filtered modes to those obtained from the corresponding emulated RIN sources. The BER values of the QD-MLL modes are very similar to the ones measured using their respective emulated RIN sources, thus validating our RIN synthesis approach.
6. Performance for different RIN profiles
The RIN emulation process proposed and validated in the previous section is now used to assess how NRZ and Manchester encoding behave when single-ended detection is used, and then scale the benefits brought by the use of Manchester encoding together with balanced detection by analysing the BER performance improvement for different RIN profiles.
The RIN profiles used in the single-ended detection case are shown in Fig. 10(a). While sharing the same low frequency part, the first three emulated RIN profiles exhibit different high-frequency constant RIN levels considered as low (Profile 1), medium (Profile 2) and high (Profile 3). The fourth RIN profile (Profile 4) is obtained from the second one by suppressing its low frequency part so that a constant RIN level is ensured over the entire synthesis bandwidth.
Fig. 10 (a) Different emulated RIN profiles considered in the single-ended detection case. (b) Comparison of Manchester encoding (MAN) and NRZ when both employ single-ended detection for the different RIN profiles.
The corresponding BER curves shown in Fig. 10(b) confirm that Manchester encoding is unable to efficiently mitigate the MPN effect when single-ended detection is used. The BER values for Manchester and NRZ are relatively close for any given RIN profile and more particularly for the high constant RIN values. Both modulation formats exhibit BER floors for RIN profiles 2 and 3. The best performance is obtained for the constant RIN profile 4 for both modulation formats. This shows that in both cases, it is the low frequency part of the RIN profile that dictates the BER performance, as can be inferred by comparing the BER curves obtained with the second and fourth RIN profiles. The frequency shift provided by Manchester encoding is not sufficient to completely avoid the overlap with the low frequency part of the RIN spectrum, which explains the absence of performance improvement brought by the use of Manchester with single-ended detection. Furthermore, the BER curves show that NRZ performs better than Manchester encoding except for Profile 3 where the BER floors of the two modulation formats appear almost at the same level. This is mainly due to the doubled modulation bandwidth required for Manchester encoding with respect to NRZ that results in more RIN to be integrated over the signal spectrum in the Manchester encoding case. Overall, the limited reduction of the overlap between the RIN and modulation spectra in the low frequency part of the spectrum is overcome by a larger amount of noise being integrated in the high frequency region as a result of the Manchester spectral shift. Considering the third RIN profile for example, the integrated RIN values over the occupied signal bandwidth are −18.3 dBc and −17.3 dBc for NRZ and Manchester coding, respectively. At this stage, it is clearly visible that the main benefit of using Manchester encoding comes from its full compatibility with balanced detection.
The different RIN levels considered in order to study the impact of the RIN profile on the performance of Manchester encoding in conjunction with balanced detection to mitigate MPN are shown in Fig. 11(a). They share the same low frequency part with different high frequency levels ranging from −140 dBc/Hz to −118 dBc/Hz. The corresponding BER curves depicted in Fig. 11(b) illustrate the effectiveness of the approach as decreasing the RIN level slightly improves the balanced detection sensitivity. The highest RIN level emulated source suffers from 4 dB power penalty at a BER of 10−9 with respect to the lowest RIN level, which provides the same performance as an ECL source. The integrated RIN over the 20-GHz occupied bandwidth for the lowest and highest RIN levels are −21.7 dBc and −16.8 dBc, respectively.
Fig. 11 (a) Different emulated RIN profiles considered in the balanced detection case. (b) BER performance obtained using Manchester encoding in conjunction with balanced detection for the different RIN profiles.
7. Conclusion
The use of Manchester encoding in conjunction with balanced detection has been shown to be an effective method to overcome the MPN-induced limitation in WDM sources making use of QD-MLLs. A clear performance improvement was experimentally demonstrated at 10 Gb/s in a 10-channel source, leading to BER enhancement and the reduction of BER floors to sub-10−9 levels. To explore in more detail the origin of the achieved improvement, use has been made of an emulated RIN source whose RIN spectrum is easily adjustable. It was shown that despite its shifted spectral content, Manchester encoding alone, i.e. with single-ended detection, is not sufficient to mitigate MPN by suppressing RIN-induced BER floors below the target BER value of 10−9. It has, however, the benefit of being compatible with balanced detection, which is largely responsible for MPN mitigation. The effectiveness of the proposed approach has been proven even at high RIN levels.
Funding
European Commission’s 7th Framework Programme (619626).
References and links
1. F. Lelarge, B. Dagens, J. Renaudier, R. Brenot, A. Accard, F. van Dijk, D. Make, O. Le Gouezigou, J.-G. Provost, F. Poingt, J. Landreau, O. Drisse, E. Derouin, B. Rousseau, P. Frederic, and G.-H. Duan, “Recent advances on InAs/InP quantum dash based semiconductor lasers and optical amplifiers operating at 1.55 μm,” IEEE J. Sel. Top. Quantum Electron. 13, 111–124 (2007). [CrossRef]
2. V. Vujicic, C. Calo, R. Watts, F. Lelarge, C. Browning, K. Merghem, A. Martinez, A. Ramdane, and L. P. Barry, “Quantum dash mode-locked lasers for data centre applications,” IEEE J. Sel. Top. Quantum Electron. 21, 1101508 (2015). [CrossRef]
3. A. Gubenko, I. Krestnikov, D. Livshtis, S. Mikhrin, A. Kovsh, L. West, C. Bornholdt, N. Grote, and A. Zhukov, “Error-free 10 Gbit/s transmission using individual Fabry-Perot modes of low-noise quantum-dot laser,” Electron. Lett. 43, 1430–1431 (2007). [CrossRef]
4. A. Akrout, A. Shen, R. Brenot, F. Van Dijk, O. Legouezigou, F. Pommereau, F. Lelarge, A. Ramdane, and G.-H. Duan, “Separate error-free transmission of eight channels at 10 Gb/s using comb generation in a quantum-dash-based mode-locked laser,” IEEE Photonics Technol. Lett. 21, 1746–1748 (2009). [CrossRef]
5. Y. Ben M’Sallem, Q. T. Le, L. Bramerie, Q.-T. Nguyen, E. Borgne, P. Besnard, A. Shen, F. Lelarge, S. LaRochelle, L. A. Rusch, and J.-C. Simon, “Quantum-dash mode-locked laser as a source for 56-Gb/s DQPSK modulation in WDM multicast applications,” IEEE Photonics Technol. Lett. 23, 453–455 (2011). [CrossRef]
6. J. Pfeifle, I. Shkarban, S. Wolf, J. N. Kemal, C. Weimann, W. Hartmann, N. Chimot, S. Joshi, K. Merghem, A. Martinez, M. Weber, A. Ramdane, F. Lelarge, W. Freude, and C. Koos, “Coherent terabit communications using a quantum-dash mode-locked laser and self-homodyne detection,” in “Optical Fiber Communication Conference,” OSA Technical Digest (online) (Optical Society of America, 2015), paper W2A.19.
7. G. Kurczveil, D. Liang, M. Fiorentino, and R. G. Beausoleil, “Robust hybrid quantum dot laser for integrated silicon photonics,” Opt. Express 24, 16167–16174 (2016). [CrossRef] [PubMed]
8. Q. Xu, B. Schmidt, J. Shakya, and M. Lipson, “Cascaded silicon micro-ring modulators for WDM optical interconnection,” Opt. Express 14, 9431–9436 (2006). [CrossRef] [PubMed]
9. C.-H. Chen, M. Ashkan Seyedi, M. Fiorentino, D. Livshits, A. Gubenko, S. Mikhrin, V. Mikhrin, and R. G. Beausoleil, “A comb laser-driven DWDM silicon photonic transmitter based on microring modulators,” Opt. Express 23, 21541–21548 (2015). [CrossRef] [PubMed]
10. M. Ashkan Seyedi, C.-H. Chen, M. Fiorentino, and R. Beausoleil, “Error-free DWDM transmission and crosstalk analysis for a silicon photonics transmitter,” Opt. Express 23, 32968–32976 (2015). [CrossRef]
11. R. Ding, Y. Liu, Q. Li, Z. Xuan, Y. Ma, Y. Yang, A. E.-J. Lim, G.-Q. Lo, K. Bergman, T. Baehr-Jones, and M. Hochberg, “A compact low-power 320-Gb/s WDM transmitter based on silicon microrings,” IEEE Photonics J. 6, 6600608 (2014). [CrossRef]
12. K. Ogawa, “Analysis of mode partition noise in laser transmission systems,” IEEE J. Quantum Electron. 18, 849–855 (1982). [CrossRef]
13. K. Iwashita and K. Nakagawa, “Suppression of mode partition noise by laser diode light injection,” IEEE J. Quantum Electron. QE-18, 1669–1982 (1982). [CrossRef]
14. M. Mielke, P. J. Delfyett, and G. A. Alphonse, “Suppression of mode partition noise in a multiwavelength semiconductor laser through hybrid mode locking,” Opt. Lett. 27, 1064–1066 (2002). [CrossRef]
15. S.-H. Yoo, S.-R. Moon, M. Kye, and C.-H. Lee, “Reduction of mode partition noise of FP-LD by using Mach-Zehnder interferometer for RSOA-based DWDM applications,” Opt. Express 24, 14494–14505 (2016). [CrossRef] [PubMed]
16. K. Sato and H. Toba, “Reduction of mode partition noise by using semiconductor optical amplifiers,” IEEE J. Sel. Top. Quantum Electron. 7, 328–333 (2001). [CrossRef]
17. M. Gay, A. O’Hare, L. Bramerie, Z. Hao, S. Fresnel, C. Peucheret, P. Besnard, S. Joshi, S. Barbet, and F. Lelarge, “Single quantum dash mode-locked laser as a comb-generator in four-channel 112 Gbit/s WDM transmission,” in Optical Fiber Communication Conference, (The Optical Society, 2014), paper Tu2H.5. [CrossRef]
18. M. E. Chaibi, L. Bramerie, S. Lobo, and C. Peucheret, “Mitigation of mode partition noise in quantum-dash Fabry-Perot mode-locked lasers using Manchester encoding,” in European Conference on Optical Communication, (2016), paper Tu.2.E.6.
19. A. Joshi, X. Wang, D. Mohr, D. Becker, and C. Wree, “Balanced photoreceivers for analog and digital fiber optic communications,” Proc. SPIE 5814, 39–50 (2005). [CrossRef]
20. Y. Yamada, Y. Shibata, T. Okugawa, and K. Habara, “High-level fluctuation tolerant optical receiver for optical packet switch and WDM cross-connect,” J. Lightwave Technol. 16, 2220–2227 (1998). [CrossRef]
21. V. Vujicic, A. P. Anthur, A. Saljoghei, V. Panapakkam, R. Zhiu, Q. Gaimard, K. Merghem, F. Lelarge, A. Ramdane, and L. P. Barry, “Mitigation of relative intensity noise of quantum dash mode-locked lasers for PAM4 based optical interconnects using encoding techniques,” Opt. Express 25, 20–29 (2017). [CrossRef] [PubMed]
22. Z. Liu, J. Xu, Q. Wang, and C. C. K. Chan, “Rayleigh noise mitigated 70-km-reach bi-directional WDM-PON with 10-Gb/s directly modulated Manchester-duobinary as downstream signal,” in “Optical Fiber Communication Conference,” (The Optical Society, 2012), paper OW1B.2.
23. J. Zhang, N. Chi, P. V. Holm-Nielsen, C. Peucheret, and P. Jeppesen, “Performance of Manchester-coded payload in an optical FSK labeling scheme,” IEEE Photonics Technol. Lett. 15, 1174–1176 (2003). [CrossRef]
24. C. M. Miller, “Intensity and noise characterizations of optical signals,” in Fiber optic test and measurement, D. Derickson, ed. (Prentice Hall, 1998).
