In this paper, we experimentally demonstrate the transmission of 56 Gbaud four-level pulse amplitude modulation (PAM4) signal over 2-km single mode fiber (SMF) with intensity modulation and direct detection (IM/DD) scheme, while the bit-error-ratio (BER) of the PAM4 signal is under hard-decision forward error correction (HD-FEC) threshold of 3.8 × 10−3. Linear pre-equalization is implemented in the transmitter side with a 3-tap finite-impulse-response (FIR) filter to compensate for the intersymbol interference (ISI) induced by system bandwidth limitation. Receiver side equalization is realized with training sequence (TS) based feed-forward equalizer (FFE) with decision-feedback equalizer (DFE). Furthermore, an Adaptive Notch Filter (ANF) is proposed to suppress the digital-to-analog converter (DAC) clock leakage induced narrowband interference for the first time, and the bandwidth of the ANF is optimized to achieve the best BER performance.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Due to the proliferation of bandwidth-hungry services, such as social media, 5G mobile front haul, cloud computing and high-division television (HDTV), capacity demands of optical communication system have experienced explosive growth in the past decade. In the long-haul transmission, optical transport of bit rate per-channel beyond 100 Gb/s is now under extensive study to sustain the traffic growth. Some key technologies, such as coherent detection, digital signal processing (DSP), higher spectrum efficiency (SE) modulation formats, and super-channel with advanced spectral shaping, are applied in the long haul transmission transponders to improve the capacity and lower cost per bit. With the popularization of datacenters, data explosion also induced the evolution in short reach optical communication systems. 400-gigabit Ethernet (400GbE) is now being discussed and standardized by IEEE task force 802.3 group, in which four optical carriers are applied and each optical carrier carries 100Gbit/s signal . Traditional electrical interconnect over coaxial cable solution cannot satisfy 400 Gbit/s signal transmission, optical interconnect becomes a good candidate for ultra-high speed signal transmission with lower cost and power consumption. Up to 2-km SMF 400 Gbit/s optical interconnect proposed for connection between transport networks and core routers as well as interconnection within datacenters has been widely discussed, low-cost solutions with intensity modulation and direction detection (IM/DD) is a good candidate in optical interconnect due to its low cost [1–6].
To achieve higher data rate signal transmission, higher SE advanced modulation formats are introduced in IM/DD system, such as pulse amplitude modulation (PAM) [3–9], Discrete Multi-Tone (DMT) [2,7–10] and carrier-less amplitude phase modulation (CAP) [7,11], are proposed and demonstrated. In addition, some linear equalization and nonlinear compensation DSP technologies are applied in the short reach to improve the performances. To realize beyond 100Gbit/s signal transmission per optical carrier with advanced modulation formats and DSP algorithms, high bandwidth Digital-to-Analog converter (DAC) and Analog-to-Digital converter (ADC). To extend the sample rate and bandwidth, analog bandwidth interleaving (ABI)  and digital bandwidth technology (DBI)  have been proposed in DAC and ADC, respectively. The key technologies in ABI and DBI are bandwidth slicing and reorganization . These can be realized by digital/analog bandwidth filtering, frequency shift and combination. During these procedures, signal distortions sometimes appear and some optimized circuits design and DSP technologies are applied to mitigate the distortions . Among the distortions, clock-leakage can always be found in high-speed DAC and ADC. In ultra-high capacity OFDM signal transmission with DAC, we first observed the clock leakage from one of the differential output ports of the 65GSa/s sample rate Fujitsu DAC . In multicarrier signal, such as OFDM and DMT, transmission scheme, clock leakage is not a serious problem as subcarriers affected by clock leakage can be reserved null to avoid the performance degradation. While in single carrier transmission, clock-leakage mitigation is first proposed in , the author observed the 20 GHz clock leakage in 80 GSa/s Fujitsu DAC. Both transmitter side and receiver side DSP are proposed and demonstrated in  for DAC clock leakage mitigation. Transmitter side clock leakage mitigation is really complicated as rigid traversal search and rigid iterations are necessary in the initial calibration, while the receiver side DSP is also unpractical as ideal FFT filter are used and the bandwidth of FFT is quite narrow compared to the sampling rate of the ADC in the Oscilloscope, which mean a really large size FFT is needed. The clock leakage has really ultra-narrow bandwidth and it can be regarded as narrowband interference in optical communication system. To supress such narrowband interference practically in high capacity optical interconnect, a 2-tap least-mean-square (LMS) based Adaptive Notch Filter (ANF) is proposed to suppress the DAC clock leakage induced narrowband interference in this paper, and the bandwidth of the proposed ANF is optimized to achieve the best BER performance.
In this paper, we experimentally demonstrate the transmission of 56 Gbaud four-level PAM (PAM4) signal IM-DD scheme, after 2-km SMF transmission the bit-error-ratio (BER) of the PAM4 signal is still under hard-decision forward error correction (HD-FEC) threshold of 3.8 × 10−3. A 3-tap finite impulse-response (FIR) is applied in the transmitter to compensate the system bandwidth limitation. Receiver side equalization is realized with training sequence (TS) based feed-forward equalizer (FFE) and decision-feedback equalizer (DFE). Furthermore, an Adaptive Notch Filter (ANF) is proposed to suppress the DAC clock leakage induced narrowband interference in the system for the first time, and the bandwidth of the designed ANF is optimized to achieve the best BER performance.
The elimination of clock leakage can be realized with a notch filter. A very narrow notch filter is usually desired in order to filter out the interference without distorting the signal. As the narrowband interference might be not stable and the bandwidth of notch filter is ultra-narrow, some components of signal may be suppressed by a fixed notch filter while leaving the interference. Thus, this may lead to an increase in the noise level. Also, there are many applications where the frequency of interfering sinusoid drifts slowly. A large bandwidth notch filter is desired in this case, while this will lead to signal distortion. In the above mentioned situations, a precise frequency interference measurement is necessary before a fixed notch filter can be applied to cancel the interference. An alternative technique of eliminating clock leakage narrowband interference is by an adaptive notch filter (ANF), in this scheme a narrowband interference correlated reference signal is needed while the precise frequency interference measurement is avoided .
Figure 1 shows the schematic diagram of LMS based ANF. We have to point out that the LMS algorithm used in the ANF is used to track the change of the interference instead of the transmitted signal. It is different from the LMS algorithm in the FFE and DFE based equalization, which is applied to track the signal distortion induced by the noise. The input signal consists of broadband signal and narrowband interference, the aim of the proposed ANF is to suppress while still maintain. A narrowband interference correlated reference signal, where and are the amplitude and phase of the reference signal and is the sampling period, is first passed through an adaptive filter and then the re-constructed is subtracted from the input signal to cancel the narrowband interference. There are two taps in the adaptive filter and the taps are updated with LMS criterion as follow:17]:
The bandwidth of ANF is jointly determined by adaptation step size, amplitude of reference signal and the sampling period . The bandwidth is chosen to filter out the narrowband interference as much as possible with introducing negligible signal distortions. Thus, different optimum bandwidths exist for different narrowband interferences and the optimum bandwidth in our experiment is obtained by adjusting appropriate adaptation step sizeand amplitude of reference signalto get the best BER performance.
3. Experimental setup
Figure 2 shows the experimental setup for 56 Gbaud PAM4 signal transmission and reception in IM-DD system. In the transmitter, an optical carrier signal at 1552.375 nm emitting from an external cavity laser (ECL) is modulated by 56 Gbaud PAM4 signal via a 30 GHz Mach-Zander Modulator (MZM). The power and linewidth of the optical carrier are 15.5 dBm and less than 100 KHz, respectively. 56 Gbaud PAM4 signal is generated off-line in Matlab and then uploaded into an 80 GSa/s Sampling rate Fujitsu DAC with 16 GHz 3-dB bandwidth and 8-bit resolution. The generated PAM4 signal from one of the differential output ports of DAC is amplified by a single-ended 30 GHz linear driver with 20-dB gain and then injected into the RF port of MZM to modulate the optical carrier. The MZM is biased at Quadrature point of MZM power transmission curve and the output power is 8.7 dBm. In the PAM4 signal generation, pseudorandom binary sequence (PRBS) is firstly generated and then mapped to PAM4 signal. Two types of training sequences (TSs) are transmitted in the front of 363924 PAM4 symbols, 1024 on-off key (OOK) and 2048 PAM4 symbols are transmitted for time synchronization and channel equalization, respectively. The total length of symbols in one frame is 366996. Nonlinear pre-distortion is applied to compensate the MZM transmission curve induced nonlinear distortion. Linear pre-equalization is also realized with a 3-tap FIR filter in the transmitter to compensate for the intra-symbol interference (ISI) induced by bandwidth limitation. Square-Root Raised Cosine (SRRC) filter with 0.1 roll-off factor is applied to realize Nyquist shaping for 56 Gbaud PAM4 signal. The Nyquist shaping PAM4 signal is down-sampled to 80 Gsa/s before uploaded into DAC. The total bit rate of PAM4 signal in the system is 112Gbit/s. The fiber length in this experiment is 2 km, optical amplifier is not necessary in experimental setup. The modulated signal is directly launched into fiber without power attenuation, as the fiber length is quite short, no fiber nonlinear effects are observed.
The optical spectra (0.01-nm resolution) of optical 56Gbaud PAM4 signal at the output of MZM are shown in Fig. 3(a). Figure 3(a) shows the optical spectra of signal at the output of MZM when the output of DAC is all set to “zero” input with black line. The 20 GHz narrowband interference can be clearly seen, and it is from the leakage of the DAC. In the proposed ANF, the frequency of the reference signal is set to 20 GHz, as the taps of ANF is updated all the time to track the slow frequency drifting (sub MHz) due to the component characteristics over time and temperature. Optical spectra of 56 Gbaud PAM4 signal without pre-equalization and with pre-equalization at the output of MZM are also given in Fig. 3(a) with blue line and red line, respectively. It can be seen that, high frequency power attenuation is compensated with pre-equalization in the transmitter. After 2-km SMF transmission, an optical tunable attenuator (ATT) is used to adjust the received optical power (ROP) of signal into the receiver. The total loss of 2-km SMF link is about 0.8 dB including coupling insertion loss. A PIN photodiode (PD) with 20 GHz 3-dB bandwidth is applied after the ATT to realize electrical conversion (O/E). The converted PAM4 signal is first boosted by a 15-dB gain pre-amplifier with 40 GHz 3-dB bandwidth and then captured by a Lecroy Oscilloscope (OSC) operating at 80 GSa/s sampling rate with 36GHz bandwidth. Finally the captured samples are fed into off-line DSP for PAM4 demodulation. The electrical spectra of received 56 Gbaud PAM4 signal without and with pre-equalization are shown in Figs. 3(b) and 3(c), respectively. In electrical domain, narrowband interference at 20 GHz can also be clearly observed. In Fig. 3(c) we can see that pre-equalization can compensate the high frequency power attenuations. The off-line DSP includes resampling, time synchronization, SRRC matching filter, ANF for narrowband interference cancellation, TSs based FFE with DFE for PAM4 signal equalization, PAM4 De-mapping and error counting. In this paper, BER was obtained by simple direct error counting with 4 frames of PAM4 signal which includes 1455696 symbols (4 × 363924 × 2 = 2911392 bits).
4. Experimental results and discussions
To verify the effective of narrowband interference cancellation with proposed ANF, we first test the 56Gbaud signal in electrical back-to-back (EBTB). One captured 56 Gbaud PAM4 signal sample in EBTB is inputted into the proposed ANF to filter out the above mentioned narrowband interference. Electrical spectra of narrowband interference canceller output after ANF and filtered narrowband interference are shown in Figs. 4(a) and 4(b), respectively. The 20 GHz narrowband interference can be easily filter out with the proposed ANF. The eye diagrams of recovered 56 Gbaud PAM4 signal without ANF and with ANF after equalization can be seen in Figs. 5(a) and 5(b), respectively. It can be seen that the eye diagram becomes clearer after narrowband interference is suppressed by the ANF.
To achieve better BER performance in OBTB, reference signal amplitude and adaptation step size in the ANF need to be adjusted to optimize the bandwidth according to the Eq. (4). BERs distribution of received 56 Gbaud PAM4 signal with −7 dBm ROP in OBTB with different reference signal amplitude and adaptation step size in ANF is tested and shown in Fig. 6. During the equalization of PAM4 signal, the taps of FFE and DFE are set at 37 and 17, respectively. The optimal area is marked out in the Fig. 6. In this area, the reference signal amplitude is about 0.25 V and adaptation step sizeis between 5 × 10−4 and 6 × 10−4. In the remaining part of this paper, the reference signal amplitude and adaptation step sizeare set to 0.25 V and 5.5 × 10−4, respectively. As the sampling rate of oscilloscope is 80 GSa/s, thus the sampling period is 12.5 fs. With such a setting, the optimum bandwidth of the proposed is achieved and the bandwidth is 5.5 MHz. This bandwidth of this ANF can also be obtained with higher amplitude and smaller adaptation step size. While smaller adaptation step size means slower convergence rate, this will lead to the BER performance degradation.
The approximate linear modulation curve in the MZM will introduce nonlinear distortion. With equal-spaced amplitude PAM4 signal ([-3 −1 1 3]), the recovered PAM4 signal with −7 dBm ROP @ OBTB is shown in Fig. 7(a), the nonlinear distortion can be seen. The space between [-3 −1] and [1,3] are less than that of [-1 1]. To solve this problem, an unequal-spaced PAM4 signal ([-3 −0.91 0.91 3]) is transmitted and a better performance recovered PAM4 signal shown in Fig. 4(b) is acquired. The unequal-spaced PAM4 signal is transmitted in the following experiment tests.
DMT signal with QPSK modulation format on each subcarrier is transmitted to acquire the channel response of our experimental system and the bandwidth of DMT signal is set to 28 GHz. The obtained channel response is inserted in Fig. 2 as an inset. It can be seen that the 3-dB bandwidth of the channel is 12 GHz, which is far less than the PAM4 signal bandwidth, thus pre-equalization is implemented in the transmitter. While the optimization of pre-equalization co-efficient will not be studied here as it is beyond the scope of this paper. BER performance of 56 Gbaud PAM4 signal versus ROP in OBTB with pre-equalization FIR filter and ANF is measured and given out in Fig. 8(a). As the bandwidths of DAC and PD are both not sufficient for 56 Gbaud PAM4 signal, the 3-tap FIR filter is applied to compensate the bandwidth limitation induced high frequency power attenuation. The taps are [-0.16 0.8 −0.16] and a 6-dB pre-compensation can be achieved. In the receiver, ANF is applied to suppress the 20 GHz narrowband interference. To verify the effect of FIR filter based pre-equalization and ANF based narrowband interference cancellation, the BERs of PAM4 signal recovered by 37 taps FFE with 17 taps DFE without pre-equalization or ANF are also given out in Fig. 8(a). From the result we can see that 4.1 dB and 1.3 dB receiver sensitivity improvements can be achieved at the HD-FEC BER threshold (3.8 × 10−3) with pre-equalization FIR filter and ANF, respectively. We then tested the BER performance of 56 Gbaud PAM4 signal under different FFE and DFE taps for the equalization, as shown in Fig. 8(b). We find that PAM4 signal recovered by 37-tap FFE with 7-tap DFE even has better performance compared with 97-tap FFE. We also measured the BER performance of PAM4 signal recovered by 37-tap FFE with 17-tap DFE and 67-tap FFE with 33-tap DFE, the BER performance can be future improved by increase the FFE taps and DFE taps. The eye diagrams of recovered 56Gbaud PAM4 signal by 17-tap FFE, 97-tap FFE and 37-tap FFE with 17-tap DFE are inserted as insets (i), (ii) and (iii) of Fig. 8(b), respectively. It can be seen that DFE equalizer brings significantly performance improvement. While at the HD-FEC BER threshold (3.8 × 10−3) the ROP of signal recovered by 37-tap FFE with 17-tap DFE is the same as the signal recovered by 67-tap FFE with 33-tap DFE. Take the computation complexity into consideration, 37-tap FFE with 17-tap DFE is used in the signal equalization.
Measured BER performance of 112 Gbit/s PAM4 versus ROP after fiber transmission is shown in Fig. 9. To see the impact of fiber chromatic dispersion (CD) of 2-km SMF, BER versus ROP in OBTB is also given out. It can be seen that no ROP penalty is observed after 2-km SMF transmission. The histogram of recovered PAM4 signal at −9 dBm ROP after 2-km SMF transmission is inserted as an inset in Fig. 9. The required ROP at BER of 3.8 × 10−3 is −12.6 dBm after 2-km SMF transmission. The launch power into the fiber in the experimental setup is 8.5 dBm, thus the optical power budget is 21.1 dB, it can be easily deployed in the application with four optical carrier 400G signal transmission within 2-km.
In conclusion, we experimentally demonstrate the transmission and reception of 56 Gbaud PAM4 signal over 2-km SMF with IM/DD, the required ROP at BER of HD-FEC threshold (3.8 × 10−3) is −12.6 dBm. Transmitter side FIR filter based linear 6-dB pre-emphasis is implemented and an LMS based 2-tap ANF is applied in receiver side to suppress the narrowband interference leaked from the clock of DAC. The optimum bandwidth is found to be 5.5 MHz in this paper. To our best knowledge, it is the first time to realize narrowband interference cancellation with a LMS-based ANF. 4.1 dB and 1.3 dB receiver sensitivity improvements can be obtained at the HD-FEC BER threshold (3.8 × 10−3) with pre-equalization FIR filter and ANF, respectively. Signal equalization is realized with TS based FFE with DFE. Take both the BER performance and computation complexity into consideration, 37-tap FFE with 17-tap DFE is used to for 112 Gbit/s PAM4 signal recovery in this paper. The total optical power budget of the proposed scheme is 21.1 dB, the scheme can be easily deployed for future 2-km four-channel 400G applications.
National Natural Science Foundation of China (NSFC) (61601199, 61575082, 61435006, 61525502, 61775085, 61490715); National High Technology 863 Research and Development of China (2015AA017102); Local Innovation and Research Teams Project of Guangdong Pearl River Talents Program (2017BT01X121); The Youth Science and Technology Innovation Talents of Guangdong (2015TQ01X606); Guangdong Provincial Natural Science Foundation(GDSF) (2015A030313328); Pearl River S&T Nova Program of Guangzhou (201710010051).
1. IEEE P802.3bs 400 Gb/s Ethernet Task Force [Online]. Available: http://www.ieee802.org/3/bm/index.html.
2. F. Li, J. Yu, Z. Cao, J. Zhang, M. Chen, and X. Li, “Experimental demonstration of four-channel WDM 560 Gbit/s 128QAM-DMT using IM/DD for 2-km optical interconnect,” J. Lightwave Technol. 35(4), 941–948 (2017). [CrossRef]
3. K. P. Zhong, W. Chen, Q. Sui, J. Man, A. P. Lau, C. Lu, and L. Zeng, “Experimental demonstration of 500Gbit/s short reach transmission employing PAM4 signal and direct detection with 25Gbps device,” in Optical Fiber Communication Conference2015, paper Th3A.3. [CrossRef]
4. N. Eiselt, H. Griesser, J. Wei, R. Hohenleitner, A. Dochhan, M. Ortsiefer, M. H. Eiselt, C. Neumeyr, J. J. V. Olmos, and I. T. Monroy, “Experimental demonstration of 84 Gb/s PAM-4 over up to 1.6 km SSMF using a 20-GHz VCSEL at 1525 nm,” J. Lightwave Technol. 35(8), 1342–1349 (2017). [CrossRef]
5. J. Lee, S. Shahramian, N. Kaneda, Y. Baeyens, J. Sinsky, L. Buhl, J. Weiner, U. Koc, A. Konczykowska, J. Dupuy, F. Jorge, R. Aroca, T. Pfau, and Y. Chen, “Demonstration of 112-Gbit/s optical transmission using 56 GBaud PAM-4 driver and clock-and-data recovery ICs,” in Eur. Conf. Opt. Commun. (2015), paper Mo.4.5.4.
6. J. Man, W. Chen, X. Song, and L. Zeng, “A low-cost 100GE optical transceiver module for 2km SMF interconnect with PAM4 modulation,” in Optical Fiber Communication Conference (2014), paper M2E.7. [CrossRef]
7. J. Shi, J. Zhang, Y. Zhou, Y. Wang, N. Chi, and J. Yu, “Transmission performance comparison for 100Gb/s PAM-4, CAP-16 and DFT-S OFDM with direct detection,” J. Lightwave Technol. 35(23), 5127–5133 (2017). [CrossRef]
8. M. H. Eiselt, N. Eiselt, and A. Dochhan, “Direct detection solutions for 100G and beyond,” in Optical Fiber Communication Conference (2017), paper Tu3I.3. [CrossRef]
9. X. Xu, E. Zhou, G. N. Liu, T. Zuo, Q. Zhong, L. Zhang, Y. Bao, X. Zhang, J. Li, and Z. Li, “Advanced modulation formats for 400-Gbps short-reach optical inter-connection,” Opt. Express 23(1), 492–500 (2015). [CrossRef] [PubMed]
10. C. Xie, P. Dong, S. Randel, D. Pilori, P. J. Winzer, S. Spiga, B. Kögel, C. Neumeyr, and M. Amann, “Single-VCSEL 100-Gb/s short-reach system using discrete multi-tone modulation and direct detection,” in Optical Fiber Communication Conference (2015), paper Tu2H.2. [CrossRef]
11. I. O. Miguel, T. Zuo, B. J. Jesper, Q. Zhong, X. Xu, and I. T. Monroy, “Towards 400GBASE 4-lane solution using direct detection of Multi-CAP signal in 14 GHz bandwidth per lane,” in Optical Fiber Communication Conference/National Fiber Optic Engineers Conference (2013), paper PDP5C.10.
12. P. Pupalaikis, B. Yamrone, R. Delbue, A. Khanna, K. Doshi, B. Bhat, and A. Sureka, “Technologies for very high bandwidth real-time oscilloscopes,” in Proc. IEEE Bipolar/BiCMOS Circuits Technol. Meet. 2014, pp. 128–135. [CrossRef]
13. C. Schmidt, V. H. Tanzil, C. Kottke, R. Freund, and V. Jungnickel, “Digital signal splitting among multiple dacs for analog bandwidth interleaving (abi),” in Proceedings of the IEEE International Conference on Electronics, Circuits, and Systems (2016). [CrossRef]
14. C. Laperle and M. O’Sullivan, “Advances in high-speed DACs, ADCs, and DSP for optical coherent transceivers,” J. Lightwave Technol. 32(4), 629–643 (2014). [CrossRef]
15. F. Li, X. Li, and J. Yu, “Performance comparison of DFT-Spread and pre-Equalization for 8 × 244.2-Gb/s PDM-16QAM-OFDM,” J. Lightwave Technol. 33(1), 227–233 (2015). [CrossRef]
16. Y. Zhu, W. Peng, Y. Cui, C. Kan, F. Zhu, and Y. Bai, “Comparative digital mitigations of DAC clock tone leakage in a single-carrier 400G system,” in Optical Fiber Communication Conference (2015), paper Th2A.17. [CrossRef]
17. B. Widrow, J. R. Glover, J. M. Mccool, J. Kaunitz, C. S. Williams, R. H. Hearn, J. R. Zeidler, E. Dong, and R. C. Goodlin, “Adaptive noise cancelling: Principles and applications,” Proc. IEEE 63(12), 1692–1716 (1975). [CrossRef]