Open Access
Add to BibSonomy
Abstract
We propose a self-homodyne system for next generation intra-datacenter networking. The proposed system has a higher spectral efficiency for the modulated signal compared to the intensity-modulation/direct-detection (IM/DD) systems and uses digital signal processing of reduced complexity compared to a conventional coherent system. The concept of the proposed system is to send the modulated signal and a tone originating from the same laser over the full-duplex fiber with the aid of circulators to be used remotely at the receiver for coherent detection. The overall system physical complexity approaches the equivalent IM/DD system giving the same target data rate for 400G systems and beyond. We experimentally demonstrate emulation of the proposed system and report data rates of 530 Gb/s, 448 Gb/s and 320 Gb/s on a single wavelength below the KP4 forward error correcting threshold over 500 m, 2 km and 10 km of single mode fiber, respectively.
© 2017 Optical Society of America
1. Introduction
Cloud computing services, online video streaming and Internet of Things (IoT) impose a huge traffic growth in intra-datacenter networking. This in turn pushes the data rate requirements of datacenter interconnects (DCIs) in intra-datacenter networking to unprecedented extent. In the last few years, 100 Gb/s Ethernet has been standardized for short reach applications over 2 km and 10 km to use four wavelengths in the 1300 nm wavelength window with 800 GHz channel spacing. Each wavelength carries 25 Gb/s on-off-keying (OOK) signal in an IM/DD system [1]. For 100 m multimode fiber links, four fiber lanes are used to enable 25 Gb/s transmission in the 850 nm wavelength window with common form factor pluggable transceivers (CFP and CFP2) [1].
In order to limit the number of used wavelengths and fiber lanes, several approaches have been proposed to increase the capacity per wavelength on a single polarization in an IM/DD system including 4-level pulse amplitude modulation (PAM-4) [2–6], discrete multi-tone technique (DMT) [7–10], and multi-band carrierless amplitude phase (multi-CAP) modulation [7, 11–13]. Advanced solutions are also proposed where polarization-division multiplexed (PDM) intensity modulation transmission is achieved using the optical signal representation in the Stokes space to realize a Stokes vector direct detection system [14, 15]. Also, a self-coherent approach for short reach applications has been proposed in [16] where a complex modulated signal using orthogonal frequency division multiplexing (OFDM) is polarization multiplexed with a tone originated with the same transmitter’s laser. The Stokes vector direct detection receiver is also adopted to retrieve the OFDM modulated signal. Most of these approaches are enabled by high speed DACs, ADCs, and digital signal processing (DSP) at the transmitter and receiver sides.
Recently, efforts are directed towards the standardization of 400G networks for short reach applications (<10 km), i.e., IEEE 400GbE Task force [17]. The consensus solution for 500 m fiber links uses four fiber lanes, where each lane operates at 50 Gbaud using PAM-4 modulation format. On the other hand, the adopted solution for 2 km and 10 km reaches utilizes WDM transmission over a single fiber using 8 wavelengths on 800 GHz grid where each wavelength carries ~25 Gbaud signal using PAM-4 modulation format. Next generation Ethernet bit rates are envisioned to be 800 Gb/s and 1.6 Tb/s transmission for short reach applications [18]. This motivates finding robust and scalable solutions that cope with the endless increase of data rate requirements of DCIs in intra-datacenter networking.
In parallel to capacity considerations while maintaining reduced system complexity, developing cost effective small form factor pluggable transceivers is desired for the massive number of DCIs in intra-datacenter networking. Thus, technology of integrated photonics is expected to achieve cost efficient small form factor solutions. Silicon photonics (SiP) and Indium Phosphide (InP) platforms are two possible hosts for integrated photonic devices, e.g., modulators and photodetectors [19, 20]. For InP modulators, experimental demonstrations have been reported for inter-datacenter networking using IQ modulation [21–26]. The integration of the InP modulators with RF drivers and lasers has been demonstrated in [26] and [25], respectively. The InP modulators have the advantage of operating at low values of Vπ with high 3-dB bandwidth relative to SiP solutions [27, 28]. On the other hand, 112 Gb/s experimental demonstration using SiP modulator for short reach applications has been reported using PAM-4 transmission on a single polarization [29]. Also, PDM transmission in a Stokes vector direct detection system has been reported using a SiP modulator and polarization multiplexed emulation [15]. In [30], SiP Stokes vector transmitter using IQ modulator polarization-multiplexed with a carrier is tested against SiP Stokes vector receiver on another chip.
In this paper, we propose a self-homodyne system for next generation intra-datacenter interconnects. Similar to conventional coherent system, this enables the full utilization of the four degrees of freedom available for modulation in optical transmission. Thus, the spectral efficiency has been increased compared to the aforementioned solutions with either two and three degrees of freedom in Stokes vector direct detection systems or one degree of freedom with multiple lanes or wavelengths in IM/DD system. In addition, this self-homodyne system has a reduced Rx-DSP complexity compared to a conventional coherent system and approaching the Stokes vector direct detection systems DSP complexity. The main concept of our proposed system is to utilize the full-duplex fibers deployed in intra-datacenter networking to connect between optical transceivers to send the tone over one fiber of the full duplex pair and the modulated signal on the other from the transmitter to the receiver for coherent reception. Optical circulators are used to launch/extract the tone and the modulated signal to/from the full-duplex fiber in order to enable peer-to-peer communication between the two transceivers connected by this full-duplex fiber. An InP DP-IQM, with 35 GHz 3-dB bandwidth and 2.5 V Vπ, is used at the transmitter side. We report for the first time to the best of our knowledge 530 Gb/s transmission over 500 m fiber on a single wavelength using 53 Gbaud PDM-32QAM signal below the KP4 FEC threshold (bit error rate (BER) of 2.2 × 10−4 [31],). In addition, we demonstrate 448 Gb/s transmission over 2 km fiber using 56 Gbaud PDM-16QAM at BER of 8.5 × 10−6 which is below the KP4 threshold by more than an order of magnitude. Finally, transmissions over 10 km at bit rates of 320 Gb/s and 360 Gb/s are achieved at symbol rates of 40 Gbaud and 45 Gbaud below BERs of 2.2 × 10−4 and 2.8 × 10−4, respectively, using 16QAM modulation format.
The rest of this paper is organized as follows: we describe the principle of the proposed self-homodyne system in section 2. Section 3 is dedicated to the experimental setup description of the system emulation. The reduced DSP stack from the conventional coherent system is discussed in section 4. The experimental results are presented and discussed in section 5. Finally, the paper is concluded in section 6.
2. Principle of the proposed self-homodyne system
The main concept of any self-homodyne system is to send the complex modulated signal and a copy of the tone originating from the same laser multiplexed in a specific domain from the transmitter (Tx) to the receiver (Rx) to be used for coherent reception. For example, the tone can be sent on a different fiber core or mode in a space-division multiplexed system [32, 33]. Another approach is to send the tone polarization-multiplexed with the modulated signal as in Stokes vector direct detection (SV-DD) system [16]. The direct advantage of the self-homodyne system is twofold: 1) the omission of the frequency offset removal and phase noise mitigation DSP blocks from the receiver DSP (Rx-DSP) in conventional coherent system deployed in long haul applications which uses a local oscillator at the receiver side, and 2) uncooled operation of the laser becomes possible which provides cost and power consumption saving for the transceiver. Both abovementioned advantages stem from the fact that the transmitted tone in a self-homodyne system replaces the traditional local oscillator (LO) at the receiver, hence, there is no longer a need for frequency and phase locking since the transmitted tone has the same central frequency and reference phase of the transmitted signal.
The proposed self-homodyne system uses the full-duplex fiber deployed in short reach applications to transmit the tone space division multiplexed with the modulated signal, i.e., by transmitting the modulated signal on one fiber of the full-duplex fiber and the tone on the other fiber. Figure 1 shows a schematic of the proposed system including two transceivers. Each coherent transmitter starts with the transmitter DSP (Tx-DSP) that will be discussed in subsection 3.2 enabled by the availability of a DAC at the transmitter. The output signal from the DAC is amplified to drive the DP-IQM. The laser is split into two branches; one branch is used to feed the DP-IQM for modulation, and the other branch is to provide the reference tone for transmission over the other lane of the full-duplex fiber. We add two circulators at each transceiver side to use the two lanes (colored in red and green in Fig. 1) of the full-duplex fiber as bidirectional fibers instead of unidirectional fibers in conventional IM/DD systems. The circulators C1 and C4 couple the modulated signal from transceiver 1 and the tone from transceiver 2, respectively, to one lane of the full-duplex fiber where they are counter-propagating. On the other hand, circulators C2 and C3 are used to couple the tone from transceiver 1 and the modulated signal from transceiver 2, respectively, to the other lane of the full-duplex fiber. For example, the modulated signal propagates on the red fiber from transceiver 1 towards transceiver 2 while its corresponding tone propagates on the green fiber from transceiver 1 to transceiver 2, while the opposite occurs from transceiver 2 to transceiver 1 where the red fiber and the green fiber are swapped. For coherent reception, circulators C1 and C2 extract the corresponding tone sent and the modulated signal transmitted from transceiver 2, respectively. Likewise, circulators C3 and C4 are used to extract the transmitted tone and the modulated signal from transceiver 1 to be forwarded to the dual polarization coherent receiver (DP-CRx) at transceiver 2. The tone path includes a polarization stabilizer to ensure tone-signal beating on both state of polarizations, i.e., to have equal LO-signal beating on both orthogonal polarizations. This polarization stabilizer can be done using integrated optoelectronics as in [34, 35]. Matched fibers are used to match the time delay between the signal and the tone to ensure the same laser phase noise at CRx to be canceled during the tone-signal beating. In addition, matching the tone and the signal paths enables the system implementation using uncooled lasers where the drift in the center frequency () of the laser affects the center frequency of the tone and the modulated signal simultaneously. This drift in the center frequency is omitted during the tone-signal beating process where the center frequency of the tone and the signal are the same after the drift (). After coherent reception, the output analog signals from the DP-CRx are converted to a digital output by the ADC to enable the Rx-DSP detailed in subsection 3.2.
Fig. 1 Schematic of the proposed self-homodyne system (PS: polarization stabilizer, MDL: matched delay line).
3. Experimental setup and DSP stack
3.1 Experimental setup
Figure 2 shows the experimental setup used to emulate the system described in the previous section in Fig. 1. To do this emulation, we connected the red fiber link shown in Fig. 1 between optical circulators C1 and C2. This provides the worst case effect of the counter propagating tone in the opposite direction of the optically modulated signal propagation direction on the same wavelength as will be discussed in the results section.
Fig. 2 Experimental setup for the proposed system emulation (MF: matching fiber, and PC: polarization controller).
The transmitter side, shown in Fig. 2, starts with an AC-coupled 8-bit DAC, operated at 84 GSps. The four differential outputs of the DAC are connected to a quad linear amplifier with 30 GHz 3-dB bandwidth. The quad linear driver generates four single-ended RF signals used to drive the fully packaged 35 GHz 3-dB bandwidth InP-based DP-IQM [23]. A 100 kHz linewidth external cavity laser (ECL) operating at 1565 nm with 15.5 dBm optical power is used as the optical source for the tone and the optical data signal. The continuous wave (CW) light from the ECL is split using 70/30 polarization maintained (PM) coupler to provide the DP-IQM with the input CW light through the 70% port.
After optical modulation, an Erbium doped fiber amplifier (EDFA) is used to boost the signal power level followed by a variable optical attenuator (VOA) which is mainly used to keep the received signal power for different symbol rates constant. The received signal power levels in the back-to-back (B2B) case, and at 500 m, 2 km, and 10 km are 6.5 dBm, 6.3 dBm, 6 dBm, and 4.5 dBm, respectively, at the input ports of the optical hybrid. The amplified signal is then launched to the SMF 28e + fiber spool through circulator C1. For the tone, the output of the 30% port of the PM coupler is connected to circulator C2 after matching the delay between paths to launch the tone to the fiber spool. Thus, the tone and the modulated optical signal are counter-propagating on the same fiber link. The fiber spools’ lengths used in our demonstration are 500 m, 2 km and 10 km which are the standard lengths used in short reach applications with full-duplex fibers. In the B2B case, we use 30 cm patch cord instead of the fiber spool shown in Fig. 2, to keep the effect of the circulators and also maintain the matched paths that include the circulators’ pigtail lengths.
At the receiver side, the tone and the modulated optical signal are extracted after propagation using circulators C1 and C2, respectively, for coherent reception. The tone and the modulated optical signal are then directed to the LO and signal input ports of the optical hybrid, respectively. The path of the tone includes a polarization controller to manually adjust the state of polarization of the tone. This replaces the polarization stabilizer in Fig. 1 which should adjust the tone’s state of polarization dynamically in the real system. The tone power levels, after the PC, equal to 7.4 dBm, 7.25 dBm, 7 dBm, and 5.4 dBm in the B2B case and at 500 m, 2 km, and 10 km, respectively. The eight outputs of the optical hybrid are connected to four balanced photodetectors (BPDs) feeding 8-bit real time oscilloscope (RTO) with four 33 GHz input channels operating at 80 GSps for offline signal processing.
It should be noted that the EDFA is used to compensate for the 10 dB loss of the optical hybrid at the receiver and the 2 dB loss of the circulators providing sufficient received optical power to the balanced photo-detectors where no transimpedance amplifiers follow. Hence, in a real system, the EDFA is not needed when high bandwidth transimpedance amplifiers (TIAs) are used after the balanced photodetection. Furthermore, we placed the EDFA at the transmitter side to avoid the induced optical noise when the input signal level is reduced by the circulators loss besides the propagation attenuation loss, if the EDFA is deployed as a pre-amplifier at the receiver.
3.2 DSP stack
Figure 3 illustrates the offline DSP stack used at the transmitter and the receiver. The transmitter DSP (Tx-DSP) starts with N-QAM symbol generation followed by pulse shaping at two samples per symbol as shown in Fig. 3(a). The pulse shaping filter is based on a root raised cosine (RRC) function with roll-off factor dependent on the operating symbol rate. The roll-factor values used at symbol rates of 28 Gbaud, 35 Gbaud, 40 Gbaud, 45 Gbaud, 50 Gbaud, 53 Gbaud, and 56 Gbaud are 1, 0.82, 0.6, 0.42, 0.28, 0.2, and 0.15, respectively. Next, the pulse shaped signal is resampled at DAC sampling rate of 84 GSps. After resampling, clipping and nonlinear compensation are performed where the transfer function of the used modulator [23] is inverted to ensure equally spaced QAM constellation points after optical modulation. Then, we pre-distort the signal using an experimentally optimized digital pre-emphasis filter to partially equalize the limited bandwidth of the transmitter components [23].
The receiver DSP (Rx-DSP) in the proposed self-homodyne system is a reduced version of the conventional coherent Rx-DSP where frequency offset removal and phase noise mitigation are not needed. The chromatic dispersion (CD) compensation is also omitted since short fiber links are deployed and the compensation can be included within subsequent adaptive filtering. As shown in Fig. 3(b), Rx-DSP starts with IQ imbalance compensation [36] followed by resampling where the rest of the Rx-DSP operates at double the symbol rate. It should be noted that we use the same reference clock for the DAC and the RTO. Thus, the clock recovery DSP block is not required in our demonstration. Next, synchronization is done to find the training symbols within the captured frame for initial polarization tracking. The polarization tracking is performed using 4 × 4 real valued multiple-input-multiple-output (MIMO) DSP block [21] shown in Fig. 3(b). Initially, we use training symbol least mean squares algorithm (TS-LMS) for initial polarization tracking. Then, we switch to decision directed least mean squares algorithm (DD-LMS) after convergence for steady state operation. In addition to the polarization tracking, the 4 × 4 real valued MIMO includes adaptive filter taps that (1) remove any residual inter-symbol interference (ISI), (2) include the RRC matched filtering, and (3) compensate for CD induced from short distance propagation. It is worth noting that if the operating wavelength is in the O-band window which is the case in most of the DCIs used for intra-datacenter networking, the CD effect will be minimal. Furthermore, since the DSP complexity is important for the target application, we study the required number of taps for the digital pre-emphasis in the Tx-DSP and the number of taps for the 4 × 4 MIMO DSP block in the Rx-DSP in the next section.
4. Results
Figure 4 shows the BER versus distance at different symbol rates using 16QAM modulation format. Below 10 km which is the longest distance that uses full-duplex fibers according to the IEEE 400GbE task force [17], we are able to achieve data rate of 448 Gb/s at a BER of 8.5 × 10−6 below the KP4 threshold using a symbol rate of 56 Gbaud on a single wavelength in the B2B case. Also, if we consider the hard decision forward error-correcting threshold (HD-FEC) of 3.8 × 10−3 at 10 km, 400 Gb/s payload rate after overhead removal can be achieved. For 40 Gbaud, 45 Gbaud, 50 Gbaud, and 56 Gbaud 16QAM signals, data rates of 320 Gb/s, 360 Gb/s, 400 Gb/s, and 448 Gb/s are achieved below BERs of 2.2 × 10−4, 2.8 × 10−4, 4.1 × 10−4 and 7.5 × 10−4 at 10 km, respectively.
It is noteworthy that the main reason for the system performance degradation at 10 km compared to B2B is not the ineffectiveness of CD compensation. The actual reason of this degradation is the stimulated Brillouin scattering (SBS), which is an inelastic scattering mechanism that occurs when a continuous wave (CW) light propagates in the fiber with relatively high power [37]. The backward reflection (in direction from C1 to C2) due to SBS generated by the tone propagation from C2 to C1 pollutes the signal propagating from C1 to C2 at optical frequency GHz where is the optical frequency of the CW tone in GHz. This backward reflection due to SBS is then downconverted by coherent reception to approximately 10.73 GHz (equal to the so-called Stokes or Brillouin shift [37]) affects the spectrum of the received signal due to the unwanted tone at 10.73 GHz which has ~55 MHz bandwidth (starting from deviating from the signal spectrum level until approaching the signal spectrum level again) as shown in Fig. 5. This is another important reason why we used the EDFA at the transmitter not at the receiver to avoid amplifying the SBS reflection effect. In a real system, this can be avoided when transceivers 1 and 2, as shown in Fig. 1, use different wavelengths (separated by more than the Brillouin shift of 11 GHz) such that the counter-propagating tone from transceiver 2/1 generates its back-reflected light away from the data signal from transceiver 1/2. Despite the SBS limitation, we believe that the current emulation demonstrates the proposed system especially that it represents the worst case scenario because both transceivers operate at the same wavelength for counter propagating signal and tone.
Fig. 5 56 Gbaud 16QAM electrical spectrum from the captured data at the receiver after 10 km signal propagation.
Figure 6 shows the BER versus distance at different symbol rates using 32QAM modulation format. At 500 m, BERs of 1.78 × 10−4 and 3.2 × 10−4 are achieved with symbol rates of 53 Gbaud and 56 Gbaud at 32QAM modulation format, respectively. This shows how the proposed system enables very high data rate that exceeds 0.5 Tb/s achieving 530 Gb/s on a single wavelength at 500 m. At 2 km, 500 Gb/s and 450 Gb/s are achieved at BERs of ~2.6 × 10−4 and ~1 × 10−4 using 50 Gbaud and 45 Gbaud signals with 32QAM modulation format, respectively.
It is expected that the paths of the signal and the tone are not exactly matched at the receiver. Thus, in Fig. 7, we investigate the effect of this mismatch on the system performance for a 56 Gbaud 16QAM signal at 2 km. BER versus the mismatch length curves are generated by adding patch cords at different lengths in the tone path. It can be seen that the system performance is not affected by a mismatch less than 2 m. Phase noise mitigation is required to compensate for a mismatch more than 2 m. It should be noted that the performance degradation due to path mismatch depends on the linewidth of the used laser as well as the operating symbol rate. As the laser linewidth-symbol duration product increases, the system performance becomes more sensitive to the mismatch between the tone and the signal paths. Also, another problem that arises from the path mismatch is the tolerance to laser frequency drift, especially if the target is to realize uncooled systems. However, results show that if we have few centimeters of path mismatch, the effect on the system performance will be negligible given the current laser technology.
Fig. 7 BER versus path mismatch length between the tone and the modulated signal using phase noise mitigation and without using phase noise mitigation.
Finally, we show the impact of the number of taps used for pre- and post-compensation at transmitter and receiver DSPs, respectively, in case of 56 Gbaud 16QAM signal at 2 km in Fig. 8. The digital pre-emphasis filter is initially optimized at 121 taps. In order to reduce the number of taps, we apply Kaiser Window [38] to smooth the energy of the filter taps before truncating them. It can be seen that decreasing the number of the digital pre-emphasis filter taps all the way to only 17 taps has a marginal impact on system performance. Around 20-30 taps for the receiver DSP are adequate which provide both CD and residual ISI compensation. Finally, it is worth noting that depending on the desired operating performance, we can pick a suboptimal point where we reduce the number of filter taps at the transmitter and the receiver while achieving the required performance. For example, if the required BER is 2.2 × 10−4 at 2 km, we can use several (Tx taps at 84 GSps, Rx taps at 112 GSps in case of 56 Gbaud) pairs to achieve this performance such as (25, 23), (17,33) and (9, 45).
Fig. 8 BER versus receiver number of taps used in the real valued 4 × 4 MIMO at different number of taps used in the digital pre-emphasis at the transmitter.
5. Conclusion
We proposed a self-homodyne system for next generation DCIs in intra-datacenter networking and we experimentally demonstrated an emulation for the proposed system using integrated InP DP-IQM. The proposed system has several advantages beside the direct reduction of the Rx-DSP complexity compared to conventional coherent system by omitting both the frequency offset removal block and the phase noise mitigation process. In addition, the system has the advantage of achieving high bit rates with low BERs on a single wavelength compared to direct detection systems. This reduces the number of used components, e.g., lasers, modulators, and drivers, required by the direct detection system to achieve the same target aggregate data rate with the addition of passive optical components like circulators and delay elements. For example, the consensus solution in the 400GbE task force for 2 km reach is the use of 8 wavelengths; each carrying PAM-4 modulated signal operating at a symbol rate of ~25 Gbaud (8 lanes with 8 lasers, 8 modulators, 8 drivers, 8 PDs, and associated 8 TIAs). On the other hand, the proposed system achieves 448 Gb/s at 2 km using a single wavelength operating at symbol rate of 56 Gbaud with 16QAM modulation format (1 laser, 4 drivers, 2 × 8 90° hybrid, 4 BPDs (8 PDs), and associated 4 TIAs). In addition, it is a good candidate to obtain a solution for uncooled lasers where the proposed system achieves the required data rates on a single wavelength and it is more immune to drifts in the laser frequency. These drifts in the center frequency are omitted during the tone-signal beating on the photodetectors of the BPDs since it affect the center frequency of the tone and the signal simultaneously provided that the tone and the signal paths are matched. In addition, the transmission in the C-band is possible since the compensation of the CD can be done within the adaptive filtering without a noticeable impact on the used number of taps, because the signal propagates over short distances in the SMF fiber. Finally, we report, to the best of our knowledge, record breaking transmissions of 530 Gb/s, and 448 Gb/s signals on a single wavelength over 500 m, and 2 km of SMF fibers, respectively, below KP4 threshold using integrated InP modulator. Despite the SBS effect encountered in the emulated system, data rate of 320 Gb/s is achieved after propagation over 10 km fiber below KP4 threshold.
References and links
1. “IEEE Standard for Information technology–Local and metropolitan area networks–Specific requirements–Part 3: CSMA/CD Access Method and Physical Layer Specifications Amendment 4: Media Access Control Parameters, Physical Layers, and Management Parameters for 40 Gb/s and 100 Gb/s Operation,” in IEEE Std 802.3ba-2010 (Amendment to IEEE Standard 802.3–2008), (2010), pp. 1–457.
2. M. Poulin, C. Latrasse, J. F. Gagne, Y. Painchaud, M. Cyr, C. Paquet, M. Morsy-Osman, M. Chagnon, S. Lessard, and D. V. Plant, “107 Gb/s PAM-4 transmission over 10 km using a SiP series push-pull modulator at 1310 nm,” in Proc. ECOC 2014, 21–25 Sept. 2014), Mo.4.5.3. [CrossRef]
3. H. Mardoyan, M. A. Mestre, R. Rios-Müller, A. Konczykowska, J. Renaudier, F. Jorge, B. Duval, J. Y. Dupuy, A. Ghazisaeidi, P. Jennevé, M. Achouche, and S. Bigo, “Single carrier 168-Gb/s Line-Rate PAM direct detection transmission using high-speed selector power DAC for optical interconnects,” J. Lightwave Technol. 34(7), 1593–1598 (2016). [CrossRef]
4. H. Mardoyan, M. Mestre, J. M. Estaran, F. Jorge, F. Blache, P. Angelini, A. Konczykowska, M. Riet, V. Nodjiadjim, J. Y. Dupuy, and S. Bigo, “84-, 100-, and 107-GBd PAM-4 intensity-modulation direct-detection transceiver for datacenter interconnects,” J. Lightwave Technol. 35(6), 1253–1259 (2016).
5. E. El-Fiky, M. Chagnon, M. Sowailem, A. Samani, M. Morsy-Osman, and D. V. Plant, “168-Gb/s single carrier PAM4 transmission for intra-data center optical interconnects,” IEEE Photonics Technol. Lett. 29(3), 314–317 (2017). [CrossRef]
6. M. Chagnon, M. Osman, M. Poulin, C. Latrasse, J.-F. Gagné, Y. Painchaud, C. Paquet, S. Lessard, and D. Plant, “Experimental study of 112 Gb/s short reach transmission employing PAM formats and SiP intensity modulator at 1.3 μm,” Opt. Express 22(17), 21018–21036 (2014). [CrossRef] [PubMed]
7. J. L. Wei, J. D. Ingham, D. G. Cunningham, R. V. Penty, and I. H. White, “Performance and power dissipation comparisons between 28 Gb/s NRZ, PAM, CAP and optical OFDM systems for data communication applications,” IEEE/OSA. J. Lightwave Technol. 30(20), 3273–3280 (2012). [CrossRef]
8. T. Chan, I. C. Lu, C. Jyehong, and W. I. Way, “400-Gb/s transmission over 10-km SSMF using discrete multitone and 1.3-μm EMLs,” IEEE Photonics Technol. Lett. 26(16), 1657–1660 (2014). [CrossRef]
9. T. Tanaka, M. Nishihara, T. Takahara, W. Yan, L. Li, Z. Tao, M. Matsuda, K. Takabayashi, and J. Rasmussen, “Experimental demonstration of 448-Gbps+ DMT transmission over 30-km SMF,” in Proc. OFC 2014, OSA Technical Digest (online) 9–13 Mar. 2014), M2I.5. [CrossRef]
10. W. Yan, L. Lei, L. Bo, C. Hao, T. Zhenning, T. Tanaka, T. Takahara, J. C. Rasmussen, and T. Drenski, “80 km IM-DD transmission for 100 Gb/s per lane enabled by DMT and nonlinearity management,” in Proc. OFC 2014, 9–13 Mar. 2014), M2I.4. [CrossRef]
11. M. I. Olmedo, T. Zuo, J. B. Jensen, Q. Zhong, X. Xu, S. Popov, and I. T. Monroy, ““Multiband carrierless amplitude phase modulation for high capacity optical data links,” J. Lightwave Technol. 32(4), 798–804 (2014). [CrossRef]
12. G. Stepniak, “Comparison of efficiency of N-dimensional CAP modulations,” J. Lightwave Technol. 32(14), 2516–2523 (2014). [CrossRef]
13. J. Wei, Q. Cheng, D. G. Cunningham, R. V. Penty, and I. H. White, ““100-Gb/s hybrid multiband (HMB) CAP/QAM signal transmission over a single wavelength,” IEEE/OSA,” J. Lightwave Technol. 33(2), 415–423 (2015). [CrossRef]
14. M. Chagnon, M. Morsy-Osman, D. Patel, V. Veerasubramanian, A. Samani, and D. Plant, “Digital signal processing for dual-polarization intensity and interpolarization phase modulation formats using stokes detection,” J. Lightwave Technol. 34(1), 188–195 (2016). [CrossRef]
15. M. Morsy-Osman, M. Chagnon, M. Poulin, S. Lessard, and D. V. Plant, “224-Gb/s 10-km transmission of PDM PAM-4 at 1.3 µm using a single intensity-modulated laser and a direct-detection MIMO DSP-based receiver,” J. Lightwave Technol. 33(7), 1417–1424 (2015). [CrossRef]
16. D. Che, A. Li, X. Chen, Q. Hu, Y. Wang, and W. Shieh, “Stokes vector direct detection for short-reach optical communication,” Opt. Lett. 39(11), 3110–3113 (2014). [CrossRef] [PubMed]
17. “IEEE P802.3bs 400GbE Task Force Public Area” (Mar. 2015), retrieved http://www.ieee802.org/3/bs/public/15_03/index.shtml.
18. “Ethernet Alliance Roadmap.”, retrieved 2016, http://www.ethernetalliance.org/roadmap/.
19. Y. Akulova, “Advances in integrated widely tunable coherent transmitters,” in Proc. OFC 2016, 20–24 Mar. 2016), W4H.1. [CrossRef]
20. D. Liang and J. E. Bowers, “Photonic integration: Si or InP substrates?” Electron. Lett. 45(12), 578–581 (2009). [CrossRef]
21. M. Y. S. Sowailem, T. M. Hoang, M. Morsy-Osman, M. Chagnon, M. Qiu, S. Paquet, C. Paquet, I. Woods, Q. Zhuge, O. Liboiron-Ladouceur, and D. V. Plant, “770 Gb/s PDM-32QAM coherent transmission using InP dual polarization IQ modulator,” IEEE Photonics Technol. Lett. 29(5), 442–445 (2017). [CrossRef]
22. M. Y. S. Sowailem, T. M. Hoang, M. Chagnon, M. Morsy-Osman, M. Qiu, S. Paquet, C. Paquet, I. Woods, O. Liboiron-Ladouceur, and D. V. Plant, “100G and 200G single carrier transmission over 2880 and 320 km using an InP IQ modulator and Stokes vector receiver,” Opt. Express 24(26), 30485–30493 (2016). [CrossRef] [PubMed]
23. M. Sowailem, T. Hoang, M. Morsy-Osman, M. Chagnon, D. Patel, C. Paquet, S. Paquet, I. Woods, O. Liboiron-Ladouceur, and D. Plant, “400-G single carrier 500-km transmission with an InP dual polarization IQ modulator,” IEEE Photonics Technol. Lett. 28(11), 1213–1216 (2016). [CrossRef]
24. N. Kikuchi, R. Hirai, and Y. Wakayama, “High-speed optical 64QAM signal generation using InP-based semiconductor IQ modulator,” in Proc. OFC 2014, 9–13 Mar. 2014), M2A.2. [CrossRef]
25. S. Chandrasekhar, X. Liu, P. J. Winzer, J. Simsarian, and R. Griffin, “Compact all-InP laser-vector-modulator for generation and transmission of 100-Gb/s PDM-QPSK and 200-Gb/s PDM-16-QAM,” J. Lightwave Technol. 32(4), 736–742 (2014). [CrossRef]
26. T. Tatsumi, N. Itabashi, T. Ikagawa, N. Kono, M. Seki, K. Tanaka, K. Yamaji, Y. Fujimura, K. Uesaka, and T. Nakabayashi, “A compact low-power 224-Gb/s DP-16QAM modulator module with InP-based modulator and linear driver ICs,” in Proc. OFC 2014, 9–13 Mar. 2014), Tu3H.5. [CrossRef]
27. Y. Ogiso, T. Yamada, J. Ozaki, Y. Ueda, N. Kashio, N. Kikuchi, E. Yamada, H. Mawatari, H. Tanobe, and S. Kanazawa, “Ultra-high bandwidth InP IQ modulator with 1.5 V Vπ,” in Proc. ECOC 2016, 18–22 Sept. 2016), 355–357.
28. G. Letal, K. Prosyk, R. Millett, D. Macquistan, S. Paquet, O. Thibault-Maheu, J.-F. Gagné, P.-L. Fortin, R. Dowlatshahi, and B. Rioux, “Low loss InP C-band IQ modulator with 40GHz bandwidth and 1.5 V Vπ,” in Proc. OFC 2015, 22–26 Mar. 2015), Th4E.3.
29. M. Chagnon, M. Morsy-Osman, M. Poulin, C. Paquet, S. Lessard, and D. V. Plant, “Experimental parametric study of a silicon photonic modulator enabled 112-Gb/s PAM transmission system with a DAC and ADC,” J. Lightwave Technol. 33(7), 1380–1387 (2015). [CrossRef]
30. P. Dong, X. Chen, K. Kim, S. Chandrasekhar, Y.-K. Chen, and J. H. Sinsky, “128-Gb/s 100-km transmission with direct detection using silicon photonic Stokes vector receiver and I/Q modulator,” Opt. Express 24(13), 14208–14214 (2016). [CrossRef] [PubMed]
31. “IEEE Standard for Ethernet Amendment 2: Physical Layer Specifications and Management Parameters for 100 Gb/s Operation Over Backplanes and Copper Cables,” in IEEE Std 802.3bj-2014 (Amendment to IEEE Std 802.3–2012 as amended by IEEE Std 802.3bk-2013), (2014), pp. 1–368.
32. B. J. Puttnam, R. Luis, J.-M. Delgado-Mendinueta, J. Sakaguchi, W. Klaus, Y. Awaji, N. Wada, A. Kanno, and T. Kawanishi, “High-capacity self-homodyne PDM-WDM-SDM transmission in a 19-core fiber,” Opt. Express 22(18), 21185–21191 (2014). [CrossRef] [PubMed]
33. B. J. Puttnam, J. Sakaguchi, J. M. D. Mendinueta, W. Klaus, Y. Awaji, N. Wada, A. Kanno, and T. Kawanishi, “Investigating self-homodyne coherent detection in a 19 channel space-division-multiplexed transmission link,” Opt. Express 21(2), 1561–1566 (2013). [CrossRef] [PubMed]
34. W. D. Sacher, T. Barwicz, B. J. F. Taylor, and J. K. S. Poon, “Polarization rotator-splitters in standard active silicon photonics platforms,” Opt. Express 22(4), 3777–3786 (2014). [CrossRef] [PubMed]
35. Q. Xu, L. Chen, M. G. Wood, P. Sun, and R. M. Reano, “Electrically tunable optical polarization rotation on a silicon chip using Berry’s phase,” Nat. Commun. 5, 5337 (2014). [CrossRef] [PubMed]
36. S. J. Savory, “Digital coherent optical receivers: algorithms and subsystems,” IEEE J. Sel. Top. Quantum Electron. 16(5), 1164–1179 (2010). [CrossRef]
37. G. P. Agrawal, Lightwave Technology: Telecommunication Systems (Wiley, 2005).
38. A. Oppenheim and R. Schafer, Discrete-Time Signal Processing, 2nd ed. (Prentice-Hall, New Jersey, 1999).
