Open Access
Add to BibSonomy
Abstract
The application of traditional coherent detection technology to optical access networks has been undermined due to its high complexity and high cost. In this paper, we propose a novel IQ-interleaved detection method which uses the preset frequency offset of the lasers at the transmitter and receiver to obtain the in-phase and quadrature components of the received signal. It keeps the simple structure of heterodyne detection and avoids the down-conversion process. Without Nyquist pulse shaping, the received signal bandwidth of the proposed scheme is theoretically 0.5B smaller than that of heterodyne detection for signal with a symbol rate of B. The 50-Gb/s NRZ transmission experiment proves that by using the proposed scheme, the receiving sensitivity and the frequency drift tolerance can be improved by ∼1 dB and 1 GHz compared with heterodyne detection under strong bandwidth limitation. Without pulse shaping, the receiving sensitivity, frequency drift tolerance (1-dB sensitivity penalty) and link power budget for 20-km fiber transmission are -31.8 dBm, 11 GHz and 43.5 dB, respectively. A higher power budget of 45 dB can be achieved when Nyquist pulse shaping is applied. The proposed scheme provides a low-complexity potential solution for a next-generation coherent PON.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
With the rapid growth of the numbers of smart terminals and diversified mobile technologies, user demand for network bandwidth is increasing at a rate of nearly 50% per year [1]. As a mainstream access method, high-speed optical access network has the potential to meet the above need. To facilitate operation and maintenance, most optical fiber access uses a passive optical network (PON) structure, that is, no active devices are used in the physical layer architecture between the central office and the user end. At present, commercial passive optical networks mostly adopt a tree structure, using optical splitters and time division multiplexing (TDM) [2] to achieve multi-user access. Traditional optical access networks, including E/GPON, 10GEPON, and XGPON all use on-off keying (OOK) modulation, due to the maturity of intensity modulation with direct detection (IM-DD) technology and its low cost. Recently, the channel rate of passive optical networks (PONs) has moved towards 50 Gb/s/λ and beyond, resulting in poor receiver sensitivity and serious chromatic dispersion for IM-DD [3].
Coherent detection has better performance of receiver sensitivity and higher spectral efficiency compared to IM-DD, but it comes at a greater financial cost [4] due to its high complexity of system structure and heavy digital signal processing (DSP). A standard coherent receiver for homodyne detection without polarization diversity includes a local oscillator (LO) laser, a 90-degree optical hybrid, two balanced photodiode pairs (BPDs) and two analogue-to-digital converters (ADCs). For heterodyne detection, the number of photodiode (PD) required above is halved but it requires higher receiver bandwidth and a down-conversion process in the electrical domain [5]. For cost-sensitive passive optical networks, such a high-complexity coherent detection system is obviously unacceptable.
To apply coherent detection in TDM-PON feasibly, varied coherent structures have been proposed to reduce the complexity of coherent detection system [6–12]. The receivers based on 3×3 coupler [6] have been extensively used to reduce the number of PDs in the system. A polarization-independent structure based on 3×3 coupler is further proposed to receive intensity modulated signals without DSP [7,8]. Recently, single-photodiode structure is used as a substitute for balance detection method to minimize the number of PDs in the system and a high-power local oscillator is adopted to suppress the signal-to-signal beat interference (SSBI) [9,10]. Besides, Bifrost has proposed a polarization independent quasi-coherent receiver which uses envelope detection technology to demodulate the intensity signal [11,12]. In addition to seeking structural simplification, reducing the complexity of DSP is also extremely important for coherent detection. Self-homodyne technology sends a pilot carrier at the transmitter and extract it as the local oscillator by different means at the receiver, which relaxes the requirements for carrier recovery in DSP. The pilot carrier can be transmitted with signal together through another orthogonal polarization state [13] or time division multiplexing [14,15]. And polarization-diversity self-coherent structure has been further proposed to increase the system capacity [16]. Apart from these, phase insensitive intensity modulation is also an effective way to avoid carrier recovery [17].
In this paper, we propose a novel coherent detection method which uses the preset frequency offset of the lasers at the transmitter and receiver to produce a 90-degree phase difference between every two consecutive sampling points as the in-phase component and quadrature component of the received symbol. In this way, the I- and Q- components of the signal can be received alternately, thus replacing the traditional phase diversity detection process and down-conversion process. The proposed method is experimentally demonstrated with 50-Gb/s/λ NRZ transmission system and compared with traditional heterodyne detection. The results show that without pulse shaping, the bandwidth requirement of the proposed scheme is narrower than that of heterodyne detection, which brings ∼1 dB improvement on receiver sensitivity in the case of strong bandwidth limitations. In addition, better frequency drift tolerance has been observed because it avoids the spectrum overlap that occurs during heterodyne detection. In section 4, we propose a modified equalization structure for the proposed scheme to eliminate signal distortion caused by bandwidth limitation, chromatic dispersion, and laser frequency drift. Accordingly, we believe that the proposed coherent detection scheme will make some valuable contribution to the low-cost coherent PON.
2. IQ-Interleaved detection with preset frequency offset
Homodyne detection refers to the case that the LO frequency $\omega_{LO}$ is equal to the signal carrier frequency $\omega_{s}$. The detector directly outputs the baseband electrical signal, so two branches are required to output in-phase and quadrature components of the received signal separately, as Fig. 1(a) shows. For heterodyne detection, there is a frequency offset between the transmitter laser and LO, i.e. $\omega_{IF} = \omega_{LO} - \omega_{s}$. The intermediate frequency signal output by the detector needs to be down-converted in the electrical domain and filtered to recover the received complex signal as Fig. 1(b) shows.
The proposed IQ-interleaved detector consists of a local oscillator (LO) and a pair of BPD, which is consistent with heterodyne detector. But instead of digitally/electrically down-converting the complex signal from the intermediate frequency, the real and imaginary part of the signal is obtained alternately on the first and second sampling point of each symbol, as Fig. 2(b) shows.
Figure 2 shows the principle of the detection method. In order to realize the IQ-interleaved detection, the frequency offset between the transmitter laser and local oscillator should be set as half the symbol rate, i.e. $(\omega_{s} - \omega_{LO}) = 2\pi R_{symbol}/2$. The sampling interval Ts is set to half the symbol period, i.e. 1/(2Rsymbol). In this case, the received symbol got from two consecutive sampling points can be expressed as:
Then, the $k$-th received symbol can be expressed by (1) and (3):
In this case, the complex signal can be obtained using only one pair of BPD. Due to the preset frequency offset, the proposed coherent detection structure is in the intradyne range, and the alternate reception of I- and Q- components avoids the down-conversion process. Compared with homodyne detection, both of the proposed scheme and heterodyne detection save half the numbers of PDs and ADCs. Figure 3 shows the signal spectra of three coherent detection schemes without Nyquist pulse shaping. For homodyne detection, the signal output by the detector is in the baseband and the signal bandwidth is B = Rsymbol Hz. The sampling rate of 2Rsymbol is needed to satisfy the sampling theorem.
Fig. 3. The signal spectrum under: (a) homodyne detection; (b) heterodyne detection; (c) IQ-interleaved detection
For heterodyne detection, the signal is at the intermediate frequency fIF, which equals B in order to ideally avoid spectrum aliasing, as Fig. 3(b) shows. But in actual application, the value of fIF is usually smaller than B due to the limitations of receiver bandwidth and the sampling rate of ADC. It means that heterodyne detection will inevitably be affected by spectrum aliasing. For IQ-interleaved detection, the preset frequency offset is 0.5Rsymbol and the signal bandwidth equals 1.5B as Fig. 3(c) shows. Thus, the receiver bandwidth required by the above scheme is 1.5B while the one required by heterodyne detection is 2B theoretically, which indicates that the proposed scheme would have better performance than heterodyne detection under the same bandwidth limitation. The sampling rate needs to be greater than twice the bandwidth of the received signal according to the sampling theorem, i.e. the sampling rate for heterodyne detection and IQ-interleaved should be 4Rsymbol and 3Rsymbol respectively. If Nyquist pulse shaping is adopted, the frequency spectrum of the baseband signal will be narrowed to B/2. In this case, the intermediate frequencies required by heterodyne detection and I-Q interleaved detection are the same. Their identical signal bandwidth and required sampling rate are B and 2Rsymbol respectively.
Note that as the π/2 phase difference between the two neighboring samples origins from the product of the frequency offset and the sampling interval $(\omega_{s} - \omega_{LO})TS$, the I- and Q-components obtained by the above scheme will become imbalanced when extra frequency offset $\Delta \omega ^{\prime}$ exists. Then, there will be a phase mismatch angle between the successively sampled I and Q components, which resulted in I-Q phase imbalance as (5) and (6) shows. This problem can be solved by the equalizer proposed in section 4, but it inevitably causes a certain sensitivity penalty. In addition, the preset frequency offset will produce a phase difference of 180 degrees between every two recovered symbols, which should be eliminated before demodulation. But for intensity modulated signals, the phase does not carry information and this process is avoided.
3. Experiment setup
Figure 4 depicts the experimental setup for 50-Gb/s NRZ coherent detection system based on the proposed IQ-interleaved detector with preset frequency offset. Heterodyne detection has the same system structure as IQ-interleaved detection, so we use this experiment setup to test two schemes. At the transmitter side, the 50-Gb/s NRZ signal is generated off-line in Matlab and then uploaded into an arbitrary waveform generator (AWG) running at 92 GSa/s. The generated NRZ signal is amplified to a peak-to-peak voltage of 2.5 V to drive a 40-G Mach-Zehnder modulator (MZM) for phase-insensitive intensity modulation. A tunable laser operating at 1550.0 nm with a linewidth of ∼100KHz is used as the laser source. The modulated optical signal is then amplified by an erbium doped fiber amplifier (EDFA). At the receiver side, a variable optical attenuator (VOA) and a polarization controller (PC) are placed after BTB/standard single mode fiber (SSMF) to adjust the received optical power (ROP) and the polarization state respectively. Another tunable laser with 100-kHz linewidth is used as the LO with power of 10 dBm, and the frequency offset between transmitter-side laser and LO for the proposed scheme and heterodyne detection are optimized as 25 GHz and 31 GHz, respectively. Then the signal and LO are mixed in a 50:50 optical coupler and detected by a balanced photodiode (BPD) with 43-GHz optical bandwidth. The output of BPD is amplified by a RF amplifier (RFA) and fed into a real-time digital storage oscilloscope (OSC) with 160-GSa/s sampling rate and 59-GHz bandwidth. Finally, the signal is processed offline DSP in Matlab.
Fig. 4. Experimental setup for 50-Gb/s NRZ coherent transmission system. Insets: (a) the procedure of transmitter offline DSP; (b) the receiver offline DSP of heterodyne detection; (c) the receiver offline DSP of IQ-interleaved detection.
The DSP flow at the transmitter side is shown in Fig. 4(a). A pseudo random binary sequence (PRBS) with length of 215−1 is firstly mapped to NRZ symbols. Before re-sampling to match the sampling rate of the AWG, the data passes through a root-raised cosine (RRC) filter with a roll-off factor of 0.1 if Nyquist pulse shaping is adopted. Figure 4 (b) and (c) depict the receiver-side DSPs for heterodyne detection and the proposed scheme respectively. For heterodyne detection, the received signal is down-converted, low-pass filtered and then re-sampled to two samples per symbol. Subsequently, synchronization and 31-tap T/2-spaced time domain equalization are performed. Finally, a symbol decision can be made after modulus operation because we adopt the phase-insensitive intensity modulation. For the proposed scheme, the receiver-side DSP flow is very similar to heterodyne detection, including re-sampling, serial-to-parallel conversion, synchronization, equalization and modulus operation. Among them, the sampling rate need to be resampled from 160 GSa/s to 100 GSa/s firstly which is consistent with heterodyne detection. However, the proposed scheme avoids the down-conversion process and the equalization method needs to be adjusted due to the special IQ-interleaved detection. The specific equalizer structure will be detailed in section 4.
Figure 5 depicts the signal spectra output by the homodyne detection, heterodyne detection and I-Q interleaved detection respectively without pulse shaping. The laser frequency offsets of the three schemes are set to 0 GHz, 31 GHz and 25 GHz. Since the bandwidth of the oscilloscope used is 59 GHz, the signal frequency spectrum is truncated. Note that the value of frequency offset set for heterodyne detection here is optimal in the case of strong bandwidth limitation. It can be seen that the signal of heterodyne detection suffers from a more serious bandwidth limitation caused by the receiver bandwidth than that of I-Q interleaved detection, which is consistent with the analysis in section 2.
Fig. 5. The signal spectrum under: (a) homodyne detection; (b) heterodyne detection; (c) IQ-interleaved detection
It is worth noting that a 40G MZM is used in the above experimental scheme, limited by experimental conditions. Similar system performance can also be achieved using 25G electro absorption modulator (EML) of 50G PON standard for lower cost. In addition, the above experimental setup has been adopted to verify the feasibility of the proposed scheme in 50-Gb/s NRZ transmission system, and it is also applicable to systems that use higher-order modulation formats, such as 100-Gb/s PAM4.
4. Equalization for IQ-Interleaved detection
Thanks to the optical field reconstruction of coherent detection structure, feedforward equalizer (FFE) is effective to compensate the linear impairments of the channel. But in proposed IQ-interleaved detection, each complex symbol is obtained by using two consecutive samples as its I- and Q-components, so these two components of the current symbol will be affected by different degrees of inter symbol-interference (ISI) from other symbols, resulting in severe I-Q imbalance. In response to the above points, we propose a T/2-spaced FFE based on training sequence and radial error is used to update taps, as Fig. 6 shows. In addition, this equalizer structure with two groups of independent taps can effectively solve the above mentioned I-Q phase imbalance problem caused by frequency drift.
In the proposed equalizer structure, ${\mathbf x}(2k - 1)$ and ${\mathbf x}(2k)$ are the consecutive sampling points and pass through two filters with different taps ${\mathbf w_{1}}$ and ${\mathbf w_{2}}$ respectively. Superscript T stands for the transpose of a vector. The outputs of the filters are used as the in-phase and quadrature components to recover the data symbol, as expression (9) shows.
For the adaption of FFE, least mean-square (LMS) algorithm is selected. The modified cost-function at the $k$-th iteration is of the form
where $E[.]$ indicates statistical expectation and $R$ represents the theoretical modulus of the signal. The algorithm aims to minimize the radial error by using stochastic gradient descent (SGD) method. In order to obtain a faster and better convergence performance, we adopt training sequence-based equalization in the experiment. The tap weights vector at the $k$-th iteration is expressed as:It's important to note that although the proposed equalization has two groups of taps that need to be calculated and updated independently, the taps are real-valued. These two groups of real taps correspond to the real and imaginary parts of the plural taps for conventional structure. So the algorithm complexity of the proposed modified equalization and the equalization in heterodyne detection is identical. Figure 7 depicts the constellation diagrams of the received NRZ signal under different transmission conditions at the received optical power (ROP) of -23 dBm without pulse shaping. Theoretically, the constellation of the NRZ signal obtained by coherent detection should be two rings. However, IQ imbalance caused by receiver bandwidth limitation, laser frequency drift and fiber dispersion makes it difficult to distinguish 0 and 1 levels. After performing the above-mentioned equalization, the signal constellation is recovered to two ideal rings, which indicates the effectiveness of the proposed equalization structure for IQ-interleaved detection. Besides, other commonly used blind equalization algorithms such as radius directed algorithm (RDA) [18] and Multi-modulus algorithm (MMA) [19] are also effective to eliminate the I-Q imbalance in this scheme.
Fig. 7. The NRZ constellation under different DSP for BTB and 20-km fiber transmission at ROP of -23 dBm.
5. Results and discussion
5.1 Without Nyquist pulse shaping
We firstly evaluate the system performance of IQ-interleaved detection without Nyquist pulse shaping, and compare it with heterodyne detection. In the case of the same receiver bandwidth of 43 GHz, both of the signal spectra under two detection schemes are severely restricted by the bandwidth of PD. The frequency offset between the transmitter-side laser and LO are optimized as 27 GHz and 31 GHz for the proposed scheme and heterodyne detection, respectively. It is worth noting that the optimal frequency offset for the proposed scheme is no longer the theoretical 25 GHz, which is due to the I-Q imbalance caused by severe bandwidth limitation. Figure 8(a) displays the receiver sensitivity of two detection schemes in back to back (BTB) case. At a BER threshold of 1×10−2, the received optical power (ROP) required is -31.8 dBm for IQ-interleaved detection, which is 0.8 dB better compared with heterodyne detection. This is because the laser frequency offset required by the proposed scheme is smaller than that of heterodyne detection, resulting in a narrower signal spectrum and a relatively weaker bandwidth limitation.
Fig. 8. The system performance: (a) the BER versus optical received power at BTB; (b) the sensitivity penalty versus frequency offset at BER of 1×10−2.
Then we test the frequency-drift tolerance of the proposed coherent detection scheme. The performance at BER of 1×10−2 as a function of frequency drift is shown in Fig. 8(b). For heterodyne detection, a sensitivity penalty of less than 1 dB is observed within [28 GHz, 38 GHz] frequency offset range. Note that the zero point of Y-axis in the figure is set to -31.8 dBm, which is the sensitivity of the proposed scheme, to make the performance difference between the two schemes more obvious. Apart from this, larger frequency offset would cause more severe ISI due to bandwidth limitation while smaller frequency offset would result in spectral overlap. For the proposed scheme, the frequency offset range is [24 GHz, 35 GHz], i.e., the tolerance of frequency drift is 11 GHz, which is slightly better than heterodyne detection. The reason is that, IQ-Interleaved detection will not be affected by spectrum aliasing, and its sensitivity penalty mainly comes from the I-Q imbalance caused by frequency drift, as described in section 2.
Then, we evaluate the system performance after 20-km and 40-km fiber transmission, as shown in Fig. 9(a). Thanks for the proposed equalization structure, the chromatic dispersion can be completely compensated in the electrical domain and there is almost no sensitivity penalty at the threshold of 1×10−2 compared with BTB case. We further test the total link power budget after 20-km fiber transmission. Due to the fiber nonlinearity, the BER performance at ROP of -31 dBm is getting worse when the launch power increases over 8 dBm. However, increasing the launch power is still an effective way to increase the total power budget, as Fig. 9(b) shows. The maximum link power budget of 43.50 dB is achieved when the launch power is 14 dBm. Without the EDFA, the maximum output power of the modulator is 4 dBm and the corresponding power budget is 35.8 dB.
Fig. 9. The system performance: (a) the BER versus optical received power under different transmission conditions; (b) the BER at the ROP of -31 dBm and the link power budget versus launch power for 20-km fiber transmission.
5.2 With Nyquist pulse shaping
Similarly, we verified the system performance of the proposed scheme with Nyquist pulse shaping and set the roll-off factor of the RRC filter to 0.1. The signal bandwidth is greatly reduced due to pulse shaping, so that the receiver bandwidth limitation has a weaker impact on the signal. For heterodyne detection, a smaller frequency spacing of signal to LO is needed and the optimal frequency offset is 26 GHz. For the proposed scheme, the optimal frequency offset is also 26 GHz, which slightly deviates from the theoretical value of 25 GHz.
The BER performance versus ROP for the two schemes at BTB is measured. As shown in Fig. 10(a), we can obtain almost identical sensitivity for heterodyne detection and IQ-interleaved detection. The sensitivity of the proposed scheme is -32.8 dBm. We also further tested the tolerance of the two schemes to frequency drift. As Fig. 10(b) shows, the permitted frequency drift range of the proposed scheme and heterodyne detection are [23 GHz, 30 GHz] and [24 GHz, 31 GHz] for a 1 dB penalty at the BER threshold of 1×10−2, respectively. It is worth noting that Nyquist pulse shaping weakens the tolerance of frequency drift of heterodyne detection. Because in the case of pulse shaping, the signal spectrum is narrow and concentrated, which is more severely affected by the spectrum aliasing resulting from frequency drift.
Fig. 10. The system performance: (a) the BER versus received optical power; (b) the sensitivity penalty versus frequency offset at BER of 1×10−2.
Figure 11(a) displays the sensitivity of IQ-interleaved detection after 20-km and 40-km fiber transmission. Consistent with the result when Nyquist pulse shaping is not performed, the fiber dispersion hardly affects the sensitivity. In summary, the sensitivity performance and frequency drift tolerance of IQ-interleaved detection with pulse shaping are the same as heterodyne detection. Finally, we also measure the total power budget in this case. The highest power budget is 45 dB when the launch power is 14 dBm, as Fig. 11(b) shows. And the link power budget is 36.8 dB without EDFA when the maximum output power of the modulator is 4 dBm. Compared with the case without pulse shaping, the system sensitivity and link power budget with pulse shaping have a certain improvement because the narrower signal spectrum makes the receiver bandwidth more abundant.
Fig. 11. The system performance: (a) the BER versus optical received power under different transmission conditions; (b) the BER at the ROP of -32 dBm and the link power budget versus launch power for 20-km fiber transmission.
5.3 Training cost of the equalization for IQ-Interleaved detection
In addition, we evaluate the training cost for the proposed training sequence-based equalization in I-Q interleaved detection. We test the BER and sensitivity penalty versus different lengths of the training sequence for 20-km fiber transmission. The tap number is set to 31 and the step size for equalizer tap updating is set to 0.001. As shown in Fig. 12, the BER and sensitivity penalty gradually decrease as the length of the training sequence increases. The equalizer has tended to converge when the training length reaches 3000 for both cases. In other parts of the experiment, we keep the length of the training sequence at 3000 in order to enable the equalizer to converge to achieve the best performance.
Fig. 12. The training cost for 20-km fiber transmission: (a) the BER versus training sequence length; (b) the sensitivity penalty versus training sequence length.
6. Conclusions
In this paper, we propose a new coherent detection scheme and experimentally demonstrate the high-sensitivity detection of a 50-Gb/s NRZ transmission system. The novel IQ-interleaved detection scheme uses almost the same structure as heterodyne detection but no digital/electrical frequency down-converting is needed to recover the complex signal. Besides, we propose a modified equalization structure for this scheme to effectively eliminate signal distortion caused by chromatic dispersion, laser frequency drift, and receiver bandwidth limitation. The experimental results show that, the proposed scheme provides similar performance with heterodyne detection when the modulated signal is Nyquist pulse-shaped, and a power budget as high as 45 dB is obtained. For the unshaped signal, the proposed scheme shows superior performance than heterodyne detection both in sensitivity and frequency drift tolerance aspects. Sensitivity higher than -30.8 dBm can be obtained within 11 GHz frequency drift using the proposed scheme, and the corresponding values are -30 dBm and 10 GHz for heterodyne detection case. Besides, according to the detection principle, the proposed scheme requires lower receiver bandwidth than heterodyne detection, and would have better performance in the case of strong receiver bandwidth limitations, which provides a potential solution for future high speed PON system.
Funding
Science and Technology Commission of Shanghai Municipality (20511102400, 20ZR1420900); 111 Project (D20031); ZTE Research Fund.
Disclosures
The authors declare no conflicts of interest.
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. P. Miguelez, “What Applications are Driving Higher Capacity in Access?” in Optical Fiber Communication Conference (2018), paper M2B.1.
2. S. Bindhaiq, A.S.M. Supa, N. Zulkifli, A.B. Mohammad, R.Q. Shaddad, M.A. Elmagzoub, and A. Faisal, “Recent development on time and wavelength-division multiplexed passive optical network (TWDM-PON) for next-generation passive optical network stage 2 (NG-PON2) - ScienceDirect,” Optical Switching and Networking. 15, 53–66 (2015). [CrossRef]
3. X. Miao, M. Bi, J. Yu, L. Li, and W. Hu, “SVM-Modified-FFE Enabled Chirp Management for 10G DML-based 50Gb/s/λ PAM4 IM-DD PON,” in Optical Fiber Communication Conference (2019), paper M2B.5.
4. B. Corcoran, B. Foo, and J. Lowery, “Single-photodiode per polarization receiver with signal-signal beat interference suppression through heterodyne detection,” Opt. Express 26(3), 3075–3086 (2018). [CrossRef]
5. K. Kikuchi, “Fundamentals of Coherent Optical Fiber Communications,” J. Lightwave Technol. 34(1), 157–179 (2016). [CrossRef]
6. M. S. Faruk, D. J. Ives, and S. J. Savory, “Technology Requirements for an Alamouti-Coded 100 Gb/s Digital Coherent Receiver Using 3 × 3 Couplers for Passive Optical Networks,” IEEE Photonics J. 10(1), 1–13 (2018). [CrossRef]
7. E. Ciaramella, “Assessment of a Polarization-Independent DSP-Free Coherent Receiver for Intensity-Modulated Signals,” J. Lightwave Technol. 38(3), 676–683 (2020). [CrossRef]
8. F. Bottoni, M. Rannello, M. Artiglia, M. Presi, and E. Ciaramella, “Coherent PON system with high-sensitivity polarization-independent receiver and no ADC/DSP,” in European Conference on Optical Communication (2015), paper Th.1.3.
9. H. Li, M. Luo, X. Li, and S. Yu, “Demonstration of 50-Gb/s/λ PAM-4 PON with Single-PD using Polarization-Insensitive and SSBI Suppressed Heterodyne Coherent Detection,” in Optical Fiber Communication Conference (2020), paper Th1B.4.
10. Y. Zhu, K. Zou, X. Ruan, and F. Zhang, “Single Carrier 400G Transmission With Single-Ended Heterodyne Detection,” IEEE Photon. Technol. Lett 29(21), 1788–1791 (2017). [CrossRef]
11. B. Cimoli, G. S. Valdecasa, A. B. Granja, J. B. Jensen, I. T. Monroy, T. K. Johansen, and J. J. V. Olmos, “An ultra-wideband schottky diode based envelope detector for 2.5 Gbps signals,” in European Microwave Conference (2016), pp.277–280.
12. J. A. Altabas, G. S. Valdecasa, M. Didriksen, J. A. Lazaro, I. Garces, I. T. Monroy, and J. B. Jensen, “Real-time 10Gbps Polarization Independent Quasicoherent Receiver for NG-PON2 Access Networks,” in Optical Fiber Communication Conference (2018), paper Th1A.3.
13. M. Nakamura, Y. Kamio, and T. Miyazaki, “Linewidth-tolerant 10-Gbit/s 16-QAM transmission using a pilot-carrier based phase-noise cancelling technique,” Opt. Express 16(14), 10611–10616 (2008). [CrossRef]
14. M. Nakamura, N. Sumimoto, Y. Takanashi, and R. Nakamura, “SPM and phase-noise tolerant optical self-homodyne using a polarization-multiplexed and intensity-modulated pilot-carrier,” IEICE ComEX 8(8), 287–292 (2019). [CrossRef]
15. D. Che, X. Chen, J. He, A. Li, and W. Shieh, “102.4-Gb/s single-polarization direct-detection reception using signal carrier interleaved optical OFDM,” in Optical Fiber Communication Conference (2014), paper Tu3G.7.
16. H. Ji, X. Zhou, C. Sun, and W. Shieh, “Polarization-diversity receiver using remotely delivered local oscillator without optical polarization control,” Opt. Express 28(15), 22882–22890 (2020). [CrossRef]
17. J. Zhang, J. S. Wey, J. Shi, and J. Yu, “Single-Wavelength 100-Gb/s PAM-4 TDM-PON Achieving Over 32-dB Power Budget Using Simplified and Phase Insensitive Coherent Detection,” in European Conference on Optical Communication (2018), paper Tu1B.1.
18. X. Xu, B. Châtelain, and D. V. Plant, “Decision Directed Least Radius Distance Algorithm for Blind Equalization in a Dual-polarization 16-QAM System,” in Optical Fiber Communication Conference (2012), paper OM2H.5.
19. W. Zhou, P. Gou, K. Wang, M. Kong, X. Li, L. Zhao, Z. Zhu, and J. Yu, “PAM-4 Wireless Transmission based on Look-up-table Pre-distortion and CMMA Equalization at V-band,” in Optical Fiber Communication Conference (2018), paper Th2A.60.
