您当前的位置： 首页 > 资源详细信息
网页快照

来源机构： 光学快报 当期论文 Apr 8, 2021 276.5KB 科技论文

 机构 (1) (2) (7) (6) (2) (1) (20) (5) (2) (1) (1) (1) (1) (1) (29) 人物 (19) (4)

## OSA | Simple SSMF-based two-channel MGDM system operating in the 0.8?μm wavelength region

Recently, there has been much interest in utilizing the polarization-division-multiplexing (PDM) technique, even in the short-reach optical interconnects implemented with direct-detection (DD) receivers [ 1 , 2 ]. This is because we can double the transmission capacity of the standard single-mode fiber (SSMF) by using this technique. However, it appears that this technique is not cost-effective enough for use in such short-reach applications due to the expensive Stokes vector receiver, which is typically made of a polarization beam splitter, a 90° optical hybrid, and four photodetectors (PDs). We note that, in the case of using the 0.8??m region, the transmission capacity of the SSMF can also be doubled by using the mode-group-division-multiplexing (MGDM) technique [ 3 , 4 ]. This is of course because the SSMF can support two mode groups (i.e.,? ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ ) in this wavelength region. In fact, the SSMF-based two-channel MGDM system seems to be well suited for the use in the short-reach optical interconnects, since it can be implemented by using the DD receivers instead of the complicate coherent receivers [ 3 – 7 ]. This is due to the small crosstalk between the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ mode groups. However, in the conventional (i.e.,?previously reported) MGDM technique, three PDs are used for the detection of two modes to avoid the large power fluctuation caused by the intra-mode group coupling (between the ${{\rm{LP}}_{11a}}$ and ${{\rm{LP}}_{11b}}$ modes) [ 7 ]. In this Letter, we implement the SSMF-based two-channel MGDM system operating in the 0.8??m region by using only two PDs. At the receiver, these two PDs are used for the detection of the signal carried by the ${{\rm{LP}}_{01}}$ mode and the multiplexed signal (i.e.,?the combined signal carried by both ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes), respectively. We then extract the signal carried by the ${{\rm{LP}}_{11}}$ mode by using a ${{2}} \times {{1}}$ multiple-input single-output (MISO) equalizer. To point out the advantages of the proposed MGDM system, we first compare its performances with those of the conventional MGDM system implemented by utilizing three PDs [ 7 ] by numerical simulations. Despite the enhanced simplicity of the proposed MGDM system, no significant difference is observed in their estimated performances. We then experimentally demonstrate the transmission of the ${{2}} \times {{28}}\;{\rm{Gb}}/{\rm{s}}$ MGDM on-off keying (OOK) signal operating at 852.6?nm over 2.2?km of the SSMF by using the proposed technique. The results show that we can achieve the bit-error rate (BER) of ${\lt}{3.8} \times {{10}^{- 3}}$ , even after the transmission over 2.2?km of the SSMF. Figure? 1 (a) shows a schematic diagram of the SSMF-based two-channel MGDM system implemented by using the proposed technique (i.e.,?with only two PDs). We utilize two transmitters operating in the 0.8??m region (i.e.,?TX1 and TX2) for the generation of two independent signals. These two signals are transmitted through the SSMF link via the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes, after being multiplexed by using a mode-multiplexer (M-MUX). At the receiver, in principle, it may be possible to extract the signal carried by the ${{\rm{LP}}_{01}}$ mode from the fiber (for example, by coupling with a propagation-constant-matched waveguide), so that only the signal carried by the ${{\rm{LP}}_{11}}$ mode group remains in the fiber. In this case, we can detect those two transmitted signals by using only two PDs. However, such a convenient mode-demultiplexer (M-DMUX) is not available yet. Thus, we assume implementation of the receiver by using a simple 3?dB coupler and two PDs. Accordingly, at the receiver, the transmitted MGDM signal is split into two parts by using a 3?dB coupler. One part is sent to the PD after passing through the mode filter (which is used to filter out the signal carried by the ${{\rm{LP}}_{11}}$ mode) for the detection of the signal carried by the ${{\rm{LP}}_{01}}$ mode. On the other hand, the other part is directed to PD2 for the detection of the still multiplexed signal. The output signals from PD1 and PD2 are then processed by the MISO equalizer for the extraction of the signal carried by the ${{\rm{LP}}_{11}}$ mode. Since the signal detected by PD2 can be considered as a linear combination of the two signals carried by the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes (due to the small crosstalk between two mode groups), the signal carried by the ${{\rm{LP}}_{11}}$ mode can be extracted by using a simple MISO equalizer (which is used to subtract the signal detected by PD1 from that detected by PD2). The MISO equalizer is implemented by utilizing a decision unit for the signal detected by PD1 (DEC1), a symbol-spaced feed-forward equalizer (FFE), ${h_{21}}$ , and a half-symbol-spaced FFE, ${h_{22}}$ . It should be noted that, due to the use of DEC1 in this MISO equalizer, the signal carried by the ${{\rm{LP}}_{11}}$ mode can be extracted without the effects of the noises of PD1 (which, in turn, improves the receiver sensitivity of the signal carried by the ${{\rm{LP}}_{11}}$ mode). The FFEs (i.e.,? ${h_{21}}$ and ${h_{22}}$ ) are used for the synchronization and power adjustment between the signals detected by PD1 and PD2. Thus, in the case of transmitting the signals over a long distance, it would be necessary to increase the tap length of ${h_{22}}$ to cope with the increased differential group delay (DGD) between the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes. However, we can minimize this increase of the tap length of ${h_{22}}$ by roughly adjusting the optical and electrical path lengths of PD1 and PD2. It should be noted that, if necessary, we can also utilize an additional FFE, ${h_{11}}$ , between PD1 and DEC1 to mitigate the effects of the bandwidth limitations. In comparison, Fig.? 1 (b) ? Fig. 1. Schematic diagrams of the SSMF-based two-channel MGDM systems implemented by using (a)?the proposed technique and (b)?the conventional technique [ 7 ]. Download Full Size PPT Slide PDF shows the schematic diagram of the SSMF-based two-channel MGDM system implemented by using the conventional technique (i.e.,?with three PDs) [ 7 ]. In this system, the transmitted MGDM signal is first demultiplexed by using the M-DMUX into three parts (i.e.,? ${{\rm{LP}}_{01}}$ , ${{\rm{LP}}_{11a}}$ , and ${{\rm{LP}}_{11b}}$ modes); then the demultiplexed signals are detected by using three PDs (i.e.,?PD1’, PD2’, and PD3’). Thus, for the detection of the signal carried by the ${{\rm{LP}}_{11}}$ mode group, it is needed to combine the output signals of PD2’ and PD3’ synchronously. In other words, the configuration of the receiver used for the detection of the signal carried by the ${{\rm{LP}}_{11}}$ mode group should be similar to that of the balanced receiver, as it requires the matched optical and electrical path lengths for PD2’ and PD3’. We first compare the performance of the proposed two-channel MGDM system with that of the conventional system shown in Fig.? 1 (b) by numerical simulations. For this purpose, we assume that (1)?two 28 Gbaud signals [modulated in either OOK or four-level pulse-amplitude modulation (PAM4) format] are generated by two Mach–Zehnder modulators (MZMs) having a 3?dB bandwidth of 28?GHz, (2)?the peak-to-peak voltage of the driving signals for these MZMs is set to be ${0.6}\;{{\rm{V}}_\pi}$ ( ${{\rm{V}}_\pi}$ : half-wave voltage), (3)?the 28 Gbaud signals are mode-(de)multiplexed by using the nearly ideal M-MUX and M-DMUX having a small insertion loss and no crosstalk (i.e.,?the insertion losses for the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes are as small as 0.5 and 1.0?dB, respectively) [ 8 ], (4)?the MGDM signal is detected by using the DD receivers made of PIN PDs (responsivity: 1A/W, 3?dB bandwidth: 28?GHz), and (5)?the noise figure of the electrical amplifiers used at the receiver is 6?dB. We also assume that the proposed ${{2}} \times {{1}}$ MISO equalizer consists of a decision unit (DEC1), a three-tap symbol-spaced FFE ( ${h_{21}}$ ), and a six-tap half-symbol-spaced FFE ( ${h_{22}}$ ). Under these assumptions, we first consider the case of transmitting ${{2}} \times {{28}}\;{\rm{Gb}}/{\rm{s}}$ MGDM-OOK signals over the SSMF link. In the case of using the proposed MGDM system, we estimate the received optical power in front of the 3?dB coupler (since it is a part of the receiver). In this case, the receiver sensitivities of the OOK signals carried by the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes are estimated to be almost identical (i.e.,? ${-}{9.6}\;{\rm{dBm}}$ at ${\rm{BER}} = {3.8} \times {{10}^{- 3}}$ ), as shown in Fig.? 2 , despite the fact that the received signal 2 is obtained by using the received signal 1. This is because the noises of PD1 do not affect the BER performance of the signal carried by the ${{\rm{LP}}_{11}}$ mode (due to DEC1 included in the MISO equalizer). In comparison, in the case of using the conventional system [ 7 ], the receiver sensitivities of the signals carried by the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes are estimated to be ${-}{12.2}$ and ${-}{9.7}\;{\rm{dBm}}$ , respectively. Thus, the receiver sensitivity of the signal carried by the ${{\rm{LP}}_{11}}$ mode is 2.5?dB worse than that of the signal carried by the ${{\rm{LP}}_{01}}$ mode [since the performance of the signal carried by the ${{\rm{LP}}_{11}}$ mode is affected by the combined noises of PD2’ and PD3’, while the performance of the signal carried by the ${{\rm{LP}}_{01}}$ mode is affected by the noises of PD1’ only, and the received optical power of the signal carried by the ${{\rm{LP}}_{01}}$ mode is 1?dB smaller than that of the signal carried by the ${{\rm{LP}}_{11}}$ mode due to the mode-dependent losses (MDLs) of the M-MUX and M-DMUX]. We estimate that, in this 2.5?dB penalty, ${\sim}{1.5}\;{\rm{dB}}$ is caused by the increased noises, and ${\sim}{1.0}\;{\rm{dB}}$ is caused by the MDL. From these results, we conclude that the receiver sensitivity of the proposed system is almost the same as that of the conventional system [ 7 ]. However, if we assume the use of the less ideal M-(D)MUX (having larger insertion loss and crosstalk), the receiver sensitivity of the conventional system can certainly be much worse than this. In addition, in the proposed MGDM system, the complexity of the receiver is significantly reduced, as it is implemented by utilizing a simple 3?dB coupler, a mode filter, and two PDs instead of a complicate three-channel M-DMUX and three PDs (among which two of these three PDs should have the matched optical and electrical paths for the synchronous operation). We also consider the case of transmitting ${{2}} \times {{56}}\;{\rm{Gb}}/{\rm{s}}$ MGDM-PAM4 signals over the SSMF link. The results in Fig.? 2 (a) show that the proposed MISO equalizer can be used for the PAM4 signal as well, although the receiver sensitivity of the signal carried by the ${{\rm{LP}}_{11}}$ mode is deteriorated by ${\sim}{{1}}\;{\rm{dB}}$ than that of the signal carried by the ${{\rm{LP}}_{01}}$ mode due to the propagation of the decision errors [ 9 ]. Accordingly, in the case of using the PAM4 signal, the receiver sensitivity of the proposed MGDM system is also estimated to be ${\sim}{{1}}\;{\rm{dB}}$ worse than that of the conventional system [ 7 ]. ? Fig. 2. Estimated BER curves of the ${{2}} \times {{28}}\;{\rm{Gb}}/{\rm{s}}$ MGDM-OOK and ${{2}} \times {{56}}\;{\rm{Gb}}/{\rm{s}}$ MGDM-PAM4 signals in the proposed and conventional MGDM systems by numerical simulations. Download Full Size PPT Slide PDF ? Fig. 3. (a)?Experimental setup used to demonstrate the transmission of ${{2}} \times {{28}}\;{\rm{Gb}}/{\rm{s}}$ MGDM-OOK signals in the proposed system. (b)?Eye diagram of the 28?Gb/s OOK signal measured at the output of MZM. Measured eye diagrams of the multiplexed signal at the output of the M-MUX when (c)?two OOK signals were bit-synchronized and (d)?one OOK signal was delayed by one-half of the bit period from the other OOK signal. Measured eye diagrams at the outputs of (e)?PD1 and (f)?PD2 after the transmission over 2.2?km of the SSMF. Download Full Size PPT Slide PDF Figure? 3 (a) shows the experimental setup to demonstrate the proposed two-channel MGDM system. We transmitted a ${{2}} \times {{28}}\;{\rm{Gb}}/{\rm{s}}$ MGDM-OOK signal operating in the 0.8??m region over 2.2?km of the SSMF (i.e.,?SMF-28 fiber) by using the proposed MGDM technique. In this experiment, we utilized a single-frequency laser diode operating at 852.6?nm [ 10 ]. The output of this laser was coupled to 780HP fiber [ 11 ]. We first modulated the output of this laser by using a commercial MZM (designed for the operation in the 0.8??m region) for the generation of the 28?Gb/s OOK signal, as shown in Fig.? 3 (b). We believe that the proposed MGDM system can be implemented by using directly modulated vertical-cavity semiconductor lasers (VCSELs). However, an MZM was utilized in this experiment, since we did not have the high-speed ( ${\gt}{{25}}\;{\rm{Gb}}/{\rm{s}}$ ) single-mode VCSELs in our laboratory. We split the OOK signal into two parts by using a 3?dB coupler and directed to the M-MUX based on a mode-selective fiber coupler (crosstalk: ${-}{30} \;{\rm{dB}}$ ) [ 8 ]. One part was used to excite the ${{\rm{LP}}_{01}}$ mode of the SSMF (after passing through an optical delay line for the decorrelation between two OOK signals), while the other part was used to excite the ${{\rm{LP}}_{11}}$ mode. Figures? 3 (c) and 3 (d) show the measured eye diagrams of the multiplexed signals at the output of the M-MUX, when two OOK signals are bit-synchronized, and the signal carried by the ${{\rm{LP}}_{01}}$ mode is delayed by one-half of the bit period from the signal carried by the ${{\rm{LP}}_{11}}$ mode, respectively. We transmitted the multiplexed signal over the 2.2?km long SSMF link (including two fiber connectors). The fiber loss was measured to be 2.2?dB/km for both the ${{\rm{LP}}_{01}}$ and the ${{\rm{LP}}_{11}}$ modes. After the transmission, this MGDM signal was split into two parts by using a variable coupler. One output fiber of this coupler was fusion-spliced to the short patch cord made of HI780 fiber (to strip out the ${{\rm{LP}}_{11}}$ mode) [ 12 ], which was then connected to PD1 for the detection of the signal carried by the ${{\rm{LP}}_{01}}$ mode. On the other hand, the other output fiber was directed to PD2 for the detection of the multiplexed signal. However, we noted that, due to the limited resources in our laboratory, a PIN PD with no transimpedance amplifier (TIA) was used for PD1, while a PIN-TIA receiver was utilized for PD2. The 3?dB bandwidths of PD1 and PD2 were 25 and 22?GHz, respectively. Figure? 3 (e) shows the binary signal measured at the output of PD1. However, the eye diagram measured at the output of PD2 indicated a ternary signal, as shown in Fig.? 3 (f). Thus, we concluded that the signals carried by the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes were linearly combined due to the low crosstalk. For example, Fig.? 4 shows ? Fig. 4. Measured mode crosstalk after the M-MUX and variable coupler (with and without inserting the 2.2?km long SSMF). Download Full Size PPT Slide PDF the measured mode crosstalk by using an amplified-spontaneous-emission source after the M-MUX and the variable coupler (with and without inserting the 2.2?km long SSMF). The mean value of the mode crosstalk occurred at the M-MUX was measured to be as low as ${-}{30.2}\;{\rm{dB}}$ . However, this crosstalk was increased to be ${-}{14.7}$ and ${-}{18.2}\;{\rm{dB}}$ , when we measured it after the variable coupler with and without the 2.2?km long SSMF, respectively. Thus, the crosstalk was small enough for the transmission of the MGDM-OOK signal [ 3 , 7 ]. The detected signals at PD1 and PD2 were sampled by using a digital storage oscilloscope at 50?GS/s and processed off-line. We first obtained the BER of the signal carried by the ${{\rm{LP}}_{01}}$ mode (which was demultiplexed by using the HI780 fiber), and then extracted the signal carried by the ${{\rm{LP}}_{11}}$ mode by utilizing the proposed ${{2}} \times {{1}}$ MISO equalizer consisting of a decision unit, a five-tap symbol-spaced FFE ( ${h_{21}}$ ), and a 10-tap half-symbol-spaced FFE ( ${h_{22}}$ ). It should be noted that we roughly adjusted the path lengths of PD1 and PD2 before the transmission experiment to minimize the tap number of ${h_{22}}$ required for the 2.2?km long SSMF transmission (which induced the DGD of 4.3?ns). Figures? 5 (a) and 5 (b) show the measured BER curves of the 28?Gb/s OOK signals carried by the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes, respectively. In these figures, we measured the BER curves five times and plotted all the data. The slight fluctuations in the measured BER curves were attributed to the power fluctuations of the demultiplexed signals (which were caused by the interference between the signal carried by the ${{\rm{LP}}_{01}}$ mode and the small leaked signal from the ${{\rm{LP}}_{11}}$ mode and vice versa). In the back-to-back condition, we could achieve the threshold BER of ${3.8} \times {{10}^{- 3}}$ , when the received optical power was larger than ${-}{6.5}\;{\rm{dBm}}$ . The power penalty measured after the transmission over 2.2?km of SSMF was negligible. The results in Fig.? 5 also showed that the BER performance of the signal carried by the ${{\rm{LP}}_{01}}$ mode was not affected by the bit-synchronization, since it was demultiplexed simply by using a mode stripper (i.e.,?without using any information of the signal carried by the ${{\rm{LP}}_{11}}$ mode). The performance of the signal carried by the ${{\rm{LP}}_{11}}$ mode was also insensitive to the synchronization status between two OOK signals due to the use of the proposed MISO equalizer, as shown in Fig.? 5 (b). The receiver sensitivity of the signal carried by the ${{\rm{LP}}_{11}}$ mode was measured to be ${-}{6.8}\;{\rm{dBm}}$ (at ${\rm{BER}} = {3.8} \times {{10}^{- 3}}$ ). It was interesting to note that the receiver sensitivities of the signals carried by the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes were measured to be almost identical (i.e.,? ${-}{6.5}$ and ${-}{6.8}\;{\rm{dBm}}$ ), although we utilized different types of PDs for their detections. This was because the BER performance of the signal carried by the ${{\rm{LP}}_{11}}$ mode was affected by the large decision errors occurred at DEC1 (since PD1 was implemented with no TIA). We also noted that the receiver sensitivity of the signal carried by the ${{\rm{LP}}_{01}}$ mode was measured to be almost unchanged by the 2.2?km long SSMF transmission, indicating that the effect of the chromatic dispersion was negligible. Thus, we attributed the small penalty measured after the 2.2?km long SSMF transmission for the signal carried by the ${{\rm{LP}}_{11}}$ mode ( ${\sim}{0.5}\;{\rm{dB}}$ ) mostly to the imperfect operation of the MISO equalizer. ? Fig. 5. Measured BER curves of the ${{2}} \times {{28}}\;{\rm{Gb}}/{\rm{s}}$ MGDM-OOK signals carried by the (a)? ${{\rm{LP}}_{01}}$ and (b)? ${{\rm{LP}}_{11}}$ modes in the back-to-back condition and after the transmission over 2.2?km of the SSMF. Download Full Size PPT Slide PDF We have proposed a simple SSMF-based two-channel MGDM system operating in the 0.8??m wavelength region. In this system, the receiver was implemented by using a 3?dB coupler, a mode filter, two PDs, and a ${{2}} \times {{1}}$ MISO equalizer. One PD (i.e.,?PD1) was used for the detection of the signal carried by the ${{\rm{LP}}_{01}}$ mode after stripping out the signal carried by the ${{\rm{LP}}_{11}}$ mode simply by using a short length of HI780 fiber. The other PD (i.e.,?PD2) was used for the detection of the multiplexed signal (i.e.,?carried by both the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes). We then extracted the signal carried by the ${{\rm{LP}}_{11}}$ mode by using a simple ${{2}} \times {{1}}$ MISO equalizer consisting of a decision unit (used for the signal carried by the ${{\rm{LP}}_{01}}$ mode), a symbol-spaced FFE (used to process the output of this decision unit), and a half-symbol-spaced FFE (used for the equalization of the multiplexed signal). Due to this decision unit included in the MISO equalizer, we could detect the signal carried by the ${{\rm{LP}}_{11}}$ mode without the effects of the noises of PD1 (used to detect the signal carried by the ${{\rm{LP}}_{11}}$ mode). We confirmed by numerical simulations that the proposed MGDM technique could be used for the PAM4 signals, as well as the OOK signals. For a demonstration, we transmitted ${{2}} \times {{28}}\;{\rm{Gb}}/{\rm{s}}$ MGDM-OOK signals operating at 852.6?nm over 2.2?km of the SSMF. In this experiment, we generated the MGDM signal by using a mode-selective coupler as an M-MUX. The results showed that we could achieve the BER better than ${3.8} \times {{10}^{- 3}}$ for both signals carried by the ${{\rm{LP}}_{01}}$ and ${{\rm{LP}}_{11}}$ modes, even after the transmission over 2.2?km of the SSMF. We believe that, for doubling the transmission capacity of the SSMF in short-haul applications, the proposed technique has a potential to be more cost-effective than the PDM-DD technique. Funding . Institute for Information and Communications Technology Promotion (2017-0-00702). Disclosures . The authors declare no conflicts of interest. REFERENCES . 1. K. Kikuchi, Opt. Express 22 , 1971 (2014). [ CrossRef ] ? 2. D. Che and W. Shieh, J. Lightwave. Technol. 34 , 754 (2016). [ CrossRef ] ? 3. D. Garcia-Rodriguez, J. Corral, and R. Llorente, IEEE Photonics Technol. Lett. 29 , 929 (2017). [ CrossRef ] ? 4. S. H. Bae, B. G. Kim, M. S. Kim, and Y. C. Chung, OptoElectronics and Communications Conference (OECC) (2020), paper?T2–3–5. 5. S. Arik and J. Kahn, Opt. Lett. 41 , 4265 (2016). [ CrossRef ] ? 6. B. Franz and H. Bulow, IEEE Photonics Technol. Lett. 24 , 1363 (2012). [ CrossRef ] ? 7. K. Benyahya, C. Simonneau, A. Ghazisaeidi, N. Barre, P. Jian, J. Morizur, G. Labroille, M. Bigot, P. Sillard, J. Provost, H. Debregeas, J. Renaudier, and G. Charlet, J. Lightwave Technol. 36 , 355 (2018). [ CrossRef ] ? 8. K. J. Park, K. Y. Song, Y. K. Kim, J. H. Lee, and B. Y. Kim, Opt. Express 24 , 3543 (2016). [ CrossRef ] ? 9. K. Zhong, X. Zhou, J. Huo, C. Yu, C. Lu, and A. Lau, J. Lightwave Technol. 36 , 377 (2018). [ CrossRef ] ? 10. G. Steckman, W. Liu, R. Platz, D. Schroeder, C. Moser, and F. Havermeyer, IEEE J. Sel. Top. Quantum Electron. 13 , 672 (2007). [ CrossRef ] ? 11. “Nufern 780?nm select cut-off single-mode fiber,” Datasheet No. NU0016 (Coherent, 2020). 12. “Corning HI 780 & HI 780C specialty optical fibers,” Datasheet No. M0100006 (Corning, 2010). .

增加监测目标对象/主题，请 登录 直接在原文中划词！