Open Access
Add to BibSonomy
Abstract
In this paper, we present inphase-quadrature (IQ) modulation in the O-band using dual parallel Mach-Zehnder modulators on the silicon photonics platform. The detailed design of the IQ modulator (IQM) is discussed. We then report the DC and small signal characterization of the device and investigate the performance of the device in a coherent transmission system. A bit rate of 180 Gb/s with 16-QAM modulation is achieved over 20 km of single-mode fiber without any chromatic dispersion compensation. Furthermore, we demonstrate that 77 Gbaud QPSK transmission can be achieved with a low drive voltage of 3 Vpp.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The rapid increase in datacenter traffic fueled by cloud computing and multimedia applications has created a demand for high speed, power efficient, and compact transceivers for datacenter optical interconnects. The datacenter traffic is forecasted to grow three-fold in the next five years to reach 20.6 zettabytes by 2021, where more than 70% of this data traffic stays within the datacenter [1]. The currently deployed 100 Gb/s transceivers operate using 4 × 25 Gb/s on-off keying (OOK) modulation format. The next generation Ethernet optical transceivers will operate at 400 Gb/s. The recently released 400 Gb/s IEEE Ethernet standard has introduced several changes to address this increase in bit rate [2]. Pulse amplitude modulation with four levels (PAM-4) has been selected as the modulation format of choice for 400 Gb/s systems. Additionally, the use of a stronger forward error correction (FEC) code has been included in the standard. PAM-4 provides up to double the spectral efficiency of OOK, and hence requires a less extreme increase in transceiver component’s bandwidth. With the conclusion of 400 Gb/s IEEE standard, the topic of 800 Gb/s and 1.6 Tb/s datacenter optical interconnects has attracted immense interest and discussion in the community. The spectral efficiency of PAM-4 will no longer be sufficient for 800 Gb/s and beyond data rates. Therefore, to cope with such increases, other dimensions have to be exploited such as polarization and complex modulation. In the past few years the possibility of using coherent transmission in data centers has garnered significant attention and there has been considerable discussions about the feasibility and inevitability of using coherent modulation for short reach communications systems at O-band. Dual-polarization intensity modulation / direct-detection (IMDD) has been proposed and experimentally demonstrated using Stokes vector direct-detection [3,4]. However, the traditional coherent system offers higher bit rates and better sensitivity compared to Stokes vector receiver (SVR) based systems. The complexity of SVR systems is lower than coherent systems since the SVR has a coherent receiver front-end without the local oscillator however only a coherent receiver can be used for a true 4D modulation. Moreover, coherent transmission systems have been accepted for inter-data center reaches and are expected to compete with the direct detection systems for shorter reaches in the near future. The higher spectral efficiency of higher order complex modulation formats such as 16QAM and 64QAM, can enable transmission of higher bit rate while requiring less stringent bandwidth specification for individual components of the transceiver.
In parallel, silicon photonics (SiP) has recently become a popular choice for datacenter interconnects. Taking advantage of years of complementary metal oxide semiconductor (CMOS) research and development, SiP provides a low cost, high yield platform for datacenter optical interconnects [5]. Because of the high index contrast between Si and SiO2, SiP optical components enjoy a relatively smaller footprint, this results in easier integration of SiP photonic integrated circuits (PICs) which is desirable to fit in small pluggable transceiver modules such as in the quad small-form factor pluggable (QSFP) packages used in datacenter interconnects. High bandwidth silicon photonics modulators and high responsivity and high bandwidth silicon-germanium photodetectors capable of single wavelength 100 Gb/s PAM-4 transmission and detection have been recently reported [6–10]. By employing these components, SiP integrated coherent receivers (ICR) [11,12] and IQMs have been demonstrated in C-band [13–18]. More recently conventional O-band differential drive Mach-Zehnder modulators were employed for QPSK modulation, however these approaches require further modification of transmitter and achieve transmission only at low baud rates compared to utilizing IQ modulators [19]. To assess the performance of SiP IQMs for short-reach applications, there is a clear need to study O-band transmission performance using the SiP IQ modulators to generate high baud rate high order QAM formats.
In this manuscript, we present the design, analysis and characterization of an O-band silicon photonics (SiP) IQM. We also discuss a low loss resistive thermo-optic heater used for biasing the modulator. We investigate the effects of non-linearity of the SiP modulator on 16-QAM transmission. We experimentally achieve 154 Gb/s QPSK and 180 Gb/s 16-QAM transmission over 20 km of single mode fiber using a single polarization SiP modulator below the hard decision, forward error correction (HD-FEC) threshold at 3.8 × 10−3. Furthermore, we demonstrate that 77 Gbaud QPSK transmission can be achieved with a low drive voltage of 3 Vpp.
2. Device design and fabrication
In this section, we present the design and characterization of the IQM. Figure 1 shows the schematic of the IQM. The modulator was fabricated in a multi project wafer run at Institute of Micro Electronics (IME) A*STAR on a silicon-on-insulator (SOI) wafer with a 220-nm thick silicon layer, a 2 μm thick buried oxide layer, and a high resistivity 750 Ω-cm silicon substrate. The IQM consists of two-child series push-pull (SPP) travelling wave Mach-Zehnder modulators (TWMZMs) connected in parallel using a compact low loss Y splitter/combiner. We use strip waveguide S-bends to separate each arm of the MZMs after the Y splitters. The arms of the child MZMs are separated by 31 μm, while the arms of the parent MZI are separated by 305 μm. The two arms of the child MZMs are designed to be the same length, hence creating a balanced MZI. On the other hand, one arm of the parent MZI is designed to be 100 μm longer than the other arm creating an intentional imbalance between the two arms of the parent MZI. This imbalance in the parent MZI, creates an FSR, which allows the parent MZI to be biased at the intended operation point by varying the wavelength in addition to the thermo-optic heater [6]. However, it should be noted that if the child MZIs arms were also not balanced, the FSR from the child MZIs along with the parent MZI could interfere with each other making it exceptionally hard to operate the device. Hence the child MZIs were designed to have balanced arms. The phase shifter length of each child MZM is 3 mm. To prevent any current flowing along the waveguides, 2 μm long intrinsic sections are inserted along the phase shifter for every 18 μm of PN junction, creating a 90% phase shifter fill factor. This results in 2.7 mm effective phase shifter length.
Figure 2 illustrates the PN junction cross-section of the modulator. In the SPP configuration, the PN junctions on the MZM arms are connected in series and this effectively lowers the total capacitance of the two diodes. The two PN junction phase shifters are reversed biased using a single DC voltage, connected to the P + + section. The DC bias is applied through a long thin inductive wire, to separate the DC and RF signals. The lower capacitance results in lower microwave losses, allowing the SPP TWMZMs to achieve higher bandwidths [6]. The IQM is based on the SPP TWMZM modulator presented in [20]. However, the modulator design is updated for O-band operation. SPP configuration of the child MZMs provides several advantages in both the operation and the performance of the IQM. The SPP TWMZMs are operated by only one drive signal compared to the conventional dual drive TWMZMs where two drive signals are required. This significantly lowers the operation complexity of the IQM, as the modulator can be operated with two drive signals as opposed to four for the case of an IQM with conventional dual drive child MZMs. In addition, lowering the number of RF drive signals lowers the foot print of the IQM [21]. We use 400 nm wide waveguides which are slightly wider than the strictly single mode width used for O-band operation. This increase in waveguide width, lowers the optical propagation losses in the waveguides. As the fabrication process is the same as [20], the electrode design of the modulators requires minimal update. The capacitance of the PN junction is dependent on the geometry (height and width) of the waveguide, and the doping concentration of the PN junction. The height of the waveguide and doping concentrations have remained the same as in [20]. The change in waveguide width would results in a slight change in capacitance of the PN junctions. We use a commercial simulation tool to estimate the capacitance of the PN junction. For the simulation, the peak concentration of the P and N are set at 7 × 1017 cm−3 and 5 × 1017 cm−3 respectively and we assume Gaussian distribution of the dopant ions in the silicon [22]. The capacitance of the PN junction in a 400 nm wide slab waveguide is simulated to be 230 pf/m at 0 V, and 160 pf/m at −3 V which is within 1% of the values for 500 nm wide waveguide used for 1550 nm applications. As a result, the effects of the change in waveguide geometry and PN junction capacitance on the microwave loss and characteristic impedance of the travelling wave electrodes are minimal. The travelling wave electrodes of the modulator are in the symmetric coplanar stripline configuration made of aluminum alloy. The thickness of the metal layer is set to 2 μm in the foundry process, the width of ground and signal lines are set at 60 μm and the spacing between them is set to 36 μm. These width and spacing values of the electrode result in 50 Ω characteristic impedance when the PN junctions are biased at 3 V. 50 Ω terminations formed using doped silicon are place near the termination pads of the electrodes, for on chip termination. The electrodes are connected to the 50 Ω terminations by placing gold wire balls, using a wire bonder machine.
Fig. 2 The PN junction cross-section of the modulator, with dimensions: Wn + + = 7 μm, Wn + = 0.78 μm, Wn = 0.42 μm, Wp = 0.4 μm, Wp + = μm, Wp + + = 28.6 μm, Hrib = 0.22 μm, Hslab = 0.09 μm.
As shown in Fig. 1, four thermo-optic tuners are used to bias the modulator. A thermo-optic tuner is placed on each arm of the child MZMs to control the bias of each MZM individually. However only one tuner for each child MZM is electrically connected and operational. The second tuner is added to ensure that both arms of child MZMs have equal optical losses. Similarly, two tuners are placed on each arm of the parent MZM to control the bias of the parent MZM. Figures 3(a)–3(d) show the layout, image, and cross section of the thermo-optic tuners. The tuners are formed by connecting resistive segments along the waveguide in parallel [23]. Each segment consists of two resistors created by doping the silicon slab on each side of the waveguide using the N + + dopant level with peak concentration of 1 × 1020 cm−3. Each resistive segment is 29 μm long. Since the core of the waveguide is not doped, the effects on optical loss are negligible. Due to the parallel connection of the resistive segments, the overall resistance of the tuners is minimized while the length of the tuners along the waveguides are maximized which results in more power efficient operation. The parent MZM utilizes a tuner with 5 segments, while the child MZMs use tuners with 6 segments. This results in the parent tuners having a slightly higher resistance compared to the child MZMs tuners.
Fig. 3 (a) Close up top-view of the parent tuner, (b) single section of the thermo-optic tuners, (c) micro-image of the tuners and ball bonded on-chip 50 Ω termination, and (d) cross section of the thermo-optic tuners, Wrib = 0.4 μm, Wi = 2 μm, Wn + + = 1 μm.
3. Device characterization
In this section we present the DC and small signal performances of the IQM. To characterize and operate the modulator, we first need to bias the modulator properly. Commercial IQMs usually include monitor photodetectors (MPD) on each arm of the parent MZM to control the biasing of the IQM. Since, we did not include any MPD in our design, we bias the modulator by monitoring the output power. First, we couple light into the chip and monitor the output power of the IQMs. Then the parent and child MZM tuners are used to maximize the output power of the IQM by ensuring both arms of the MZM are in-phase. Next, we use one of the parent MZM’s tuners to achieve a 90° phase shift between the two arms, which translates to an output power lower than the maximum power by 3 dB. And finally, the two child MZMs are biased at null, by minimizing the output power. When both child MZMs and the parent MZM are biased at maximum the insertion loss of the IQM excluding the grating coupler and routing losses are measured to be 4.1 dB.
Figure 4(a) shows the IV curve of the child and parent MZM’s thermo-optic tuners. As seen from the slope of the IV curve, the tuner’s resistance slightly varies with applied voltage which suggests they are non-ohmic at higher voltages. The phase-change (ΔΦ) versus voltage of the tuners is shown in Fig. 4(b). The child tuners achieve a π phase-shift at 1.7 V while the parent tuners achieve a π phase-shift at 1.9 V. The power consumption of the child and parent tuners for a π phase shift are 14.3 mW and 16.2 mW, respectively, suggesting that due to their lower total resistance, and longer length the child heaters are more power efficient than the parent heaters. We encountered a small drift in the child MZMs bias points during operation of the device, this is attributed to the small separation between the arms of the child MZMs, which causes the tuners to effect both arms of the MZMs. On the other hand, the parent MZM bias was less affected by thermo-optic cross talk. Figure 4(c) shows the phase change versus DC voltage of the PN junction phase shifters of the child MZMs. The DC Vπ of each child MZM is measured to be around 12 V. As shown in Fig. 4(c), the PN junction phase shifters have a nonlinear phase change versus voltage, which results in different behavior of a SiP modulator compared to a LiNbO3 and InP modulator [25]. For higher modulation formats such as PAM4 and 16-QAM that employ multi-amplitude levels this non-linearity can result in higher error rates, therefore a non-linear compensation is used [21].
Fig. 4 (a) IV curve of the child and parent MZM’s thermo-optic tuners, (b) the phase-change (ΔΦ) versus voltage of the thermo-optic tuners, and (c) the phase shift versus voltage of MZM’s PN junction phase shifter.
Next, we characterize the small signal response of the IQM by measuring the response of each child MZM individually. This is done by first biasing the parent modulator at the top of transmission curve (i.e. both arms of parent MZM are in phase). Next, to measure the top child MZM, we bias it at quadrature point, while the bottom child MZM is biased at null using the thermo-optic tuners, this procedure is repeated for the bottom child MZM. Figure 5 presents the electro-optic S21 and electrical S11 measurements of both child MZMs for various PN junction bias voltages. The measurements are taken by terminating the travelling wave electrode on chip, by bumping on-chip 50 Ω terminations to the electrodes as shown in Fig. 3(b). Both modulators are measured to have over 30 GHz of bandwidth when the PN junction phase shifters are reversed biased at −3 volts.
4. Transmission experiment
We investigate the performance of the modulator in a coherent transmission system. Figure 6 shows the schematic of the experimental setup. A 12 dBm tunable wavelength O-band laser is used to couple light into the SiP chip through vertical grating couplers (GCs). The insertion loss of two GCs connected back-to-back is measured to be 10 dB at 1305 nm. The excess losses due to routing waveguides from GCs to the IQM are estimated to be 3.8 dB. When both child MZMs and the parent MZM are biased at maximum the insertion loss of the IQM is measured to be 4.1 dB.
Fig. 6 Schematic of the experimental setup. TDL: Tunable delay line, SMF: single mode fiber, VOA: variable optical attenuator, PC: Polarization Controller.
Two RF drive signals are generated using two 8-bit digital-to-analog convertors (DAC) channels operating at 88 GSa/s, and the drive signals are amplified using two matched 45 GHz RF amplifiers. Two RF tunable delay lines are used to compensate for any skew between the two signals caused by mismatched RF cables and adapters in the two paths. The two signals are then applied to the IQM using 50 GHz RF probes. The modulated optical signal is amplified using a praseodymium-doped fiber amplifier (PDFA) to 0 dBm average optical power. The signal is then fed into an O-band coherent receiver. Since we did not have access to balanced O-band photodetectors (PDs) we use four single ended PDs instead of two balanced photodetectors. The common mode rejection operation is instead done using digital signal processing (DSP). The coherent receiver used for this experiment was built using discrete components, however silicon photonic based integrated coherent receivers both in C band and O band have been demonstrated [12]. Additionally, waveguide coupled silicon germanium photodetectors with more than 50 GHz OE bandwidth and 1 A/W responsivity have been demonstrated in [10] and [24], which further proves the feasibility of coherent transceivers in silicon photonics platform.
Figure 7 shows the transmitter and receiver DSP stacks used in this experiment. As the target application of the O-band IQM is short reach links within a datacenter, we employ simple DSP on both the transmitter and the receiver sides. The transmitter DSP starts with QPSK or 16QAM symbol generation. The generated symbols are then pulse shaped using a gaussian filter at 2 samples per symbol followed by resampling from 2 samples per symbol to 88 GSa/s which is the DAC sampling rate. Next, digital pre-emphasis using a finite impulse response (FIR) filter is applied on the data samples for equalization of the combined frequency response of the DAC and the RF amplifiers. Finally, the amplitude levels of the electrical 16QAM signal are controlled to pre-compensate for the non-linear transfer function of the SiP IQM and achieve equidistant amplitude levels in the optical domain. Next, data samples are quantized and uploaded to the DAC memory. On the receiver side, since single ended PDs were used instead of balanced PDs, the DSP starts by common mode rejection to get the useful signal-LO beating product. This is followed by IQ power imbalance compensation to compensate for the PD + TIA responsivity and gain variations. Next, quadrature phase error correction is done to correct any imperfections in the optical 90° hybrid and ensure orthogonality of the I and Q signals. The complex received signal is then resampled from 80 GSa/s (the sampling rate of RTO) to 2 × the symbol rate and synchronization and time recovery is applied. Finally, an equalizer with 15 taps is applied to the received signal and error counting is performed. As the device is operated in O-band there is no need for chromatic dispersion compensation even for 20 km transmission and over 10 nm away from the zero-dispersion wavelength of the SMF fiber used. This was confirmed by comparing the taps spread over time for back to back and 20 km received signal.
Figures 8(a) and 8(b) show the QPSK and 16QAM constellations at 56 and 40 Gbaud after 20 km of single mode fiber, respectively. The drive voltage of the modulators was 4.5 Vpp for the 56 Gbaud QPSK constellation and 4 Vpp for 40 Gbaud 16QAM. The 16QAM electrical drive signals were intentionally attenuated at the input of the RF amplifiers in order to limit the operation of the amplifiers to their linear gain region. The launched optical power in both cases was 0 dBm. The insertion loss of the 20 km SMF-28 spool was measured to be 7.5 dB, resulting in received optical power of −7.5 dBm at the input of the coherent receiver. The LO power for all of the transmission experiments was set at 2 dBm. Figure 8(c) shows the BER versus the symbol rate for QPSK and 16QAM modulation. We achieve below HD-FEC threshold transmission of 77 Gbaud QPSK and 45 Gbaud 16QAM over 20 km of SMF without the need for chromatic dispersion compensation.
Fig. 8 (a) 56 Gbaud QPSK constellation after 20 km of SMF fiber, (b) 40 Gbaud 16QAM constellation after 20 km of SMF, and (c) BER versus the baudrate for QPSK and 16QAM modulation. The black and red horizontal dashed lines present the HD-FEC and soft decision (SD)-FEC threshold respectively.
Next, we investigate the performance of the IQM for various drive voltages. Figures 9(a) and 9(b) show the BER performance of the transmission system versus drive voltage applied to the IQM for QPSK and 16QAM respectively. We achieve below HD-FEC transmission of 77 Gbaud QPSK, with 3 Vpp which is comparable to EML or InP based TOSA’s drive voltages [25–27], while we were able to transmit 45 Gbaud 16-QAM below HD-FEC with 4 Vpp drive voltages and 3 Vpp below soft decision (SD)- FEC. It should be noted that this (180 Gb/s) is the highest achieved bitrate with 3 Vpp drive voltage using a SiP modulator. The performance of the modulator is limited by the excess on-chip routing losses and the GC losses. We expect the performance of the modulator could be further improved by only modifying the layout of the modulator and using more efficient ways of coupling the light into the silicon chip [28,29]. Next, we reduce the received signal power using a variable optical attenuator and measure the BER while the local oscillator power is kept constant at 2 dBm. The receiver sensitivity for the 77 Gbaud QPSK modulation at the HD-FEC threshold is approximately – 17 dBm of signal power. On the other hand, we successfully achieve below HD-FEC transmission of 45 Gbaud 16QAM with −11 dBm of received signal power. At the higher symbol rate, the transmission system is also limited by the PD + TIA and RTO bandwidth. The error floor is dominated by the transmitter noise. The two main contributing factors are the low drive voltage of the modulator, and high insertion loss of the transmitter due to grating coupler and routing losses. Since the modulator is driven by only a fraction of the Vπ value, the maximum extinction ratio achieved is consequently lower, in addition due to the excess routing and coupling losses on the SiP chip the signal is further attenuated.
Fig. 9 BER performance of the transmission system versus drive voltage applied to the IQM for (a) QPSK and (b) 16QAM. BER performance of the transmission system versus received optical power for (c) QPSK and (d) 16QAM.
5. Conclusion
In this manuscript, we presented an O-band SiP IQM using series push pull Mach-Zehnder modulators. We reported the detailed design of the modulator along with the thermo-optic tuners used for biasing the modulator. The biasing procedure is presented in detail and the DC and small signal characterization of the IQM is presented. We successfully achieve below HD-FEC, 180 Gb/s 16QAM and 154 Gb/s QPSK transmission over 20 km of SMF-28. Furthermore, we demonstrate that 77 Gbaud QPSK transmission can be achieved with a low drive voltage of 3 Vpp. To the best of our knowledge this is the highest achieved bitrate transmission and lowest drive voltage combination using a SiP modulator. The maximum single wavelength transmission baudrate could easily be improved by adding a second polarization which is common for coherent transmission systems. This work supports the idea that SiP based intra-datacenter coherent solutions could be a possible candidate for next generation short reach interconnects.
References
1. Cisco, “Cisco Global Cloud Index: Forecast and Methodology, 2016-2021,” white paper-c11–738085, 2016.
2. IEEE P802.3bs 400 GbE Task Force, http://www.ieee802.org/3/bs/index.html
3. E. El-Fiky, M. Osman, M. Sowailem, A. Samani, D. Patel, R. Li, M. G. Saber, Y. Wang, N. Abadia, Y. D’Mello, and D. V. Plant, “200 Gb/s transmission using a dual-polarization O-Band silicon photonic intensity modulator for Stokes vector direct detection applications,” Opt. Express 25(24), 30336–30348 (2017). [CrossRef] [PubMed]
4. 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]
5. D. Miller, “Device requirements for optical interconnects to silicon chips,” Proc. IEEE 97(7), 1166–1185 (2009). [CrossRef]
6. D. Patel, S. Ghosh, M. Chagnon, A. Samani, V. Veerasubramanian, M. Osman, and D. V. Plant, “Design, analysis, and transmission system performance of a 41 GHz silicon photonic modulator,” Opt. Express 23(11), 14263–14287 (2015). [CrossRef] [PubMed]
7. R. Ding, Y. Liu, Y. Ma, Y. Yang, Q. Li, A. Lim, G. Lo, K. Bergman, T. Baehr-Jones, and M. Hochberg, “High-speed silicon modulator with slow-wave electrodes and fully independent differential drive,” J. Lightwave Technol. 32(12), 2240–2247 (2014). [CrossRef]
8. R. Li, D. Patel, A. Samani, E. El-Fiky, Z. Xing, M. Morsy-Osman, and D. Plant, “Silicon photonic ring-assisted MZI for 50 Gb/s DAC-less and DSP-free PAM-4 transmission,” IEEE Photonics Technol. Lett. 29(12), 1046–1049 (2017). [CrossRef]
9. R. Li, D. Patel, A. Samani, E. El-Fiky, Z. Xing, M. Sowailem, Q. Zhong, and D. Plant, “An 80 Gb/s silicon photonic modulator based on the principle of overlapped resonances,” IEEE Photonics J. 9, 1–11 (2017). [CrossRef]
10. H. Chen, P. Verheyen, P. De Heyn, G. Lepage, J. De Coster, S. Balakrishnan, P. Absil, W. Yao, L. Shen, G. Roelkens, and J. Van Campenhout, “-1 V bias 67 GHz bandwidth Si-contacted germanium waveguide p-i-n photodetector for optical links at 56 Gbps and beyond,” Opt. Express 24(5), 4622–4631 (2016). [CrossRef] [PubMed]
11. C. Doerr, L. Chen, T. Nielsen, R. Aroca, L. Chen, M. Banaee, S. Azemati, G. McBrien, S. Y. Park, J. Geyer, B. Guan, B. Mikkelsen, C. Rasmussen, M. Givehchi, Z. Wang, B. Potsaid, H. C. Lee, E. Swanson, and J. G. Fujimoto, “O, E, S, C, and L band silicon photonics coherent modulator/receiver,” in Optical Fiber Communication Conference (IEEE, 2016) paper Th5C.4. [CrossRef]
12. J. Verbist, J. Zhang, B. Moeneclaey, W. Soenen, J. Van Weerdenburg, R. Van Uden, C. Okonkwo, J. Bauwelinck, G. Roelkens, and X. Yin, “A 40-GBd QPSK/16-QAM integrated silicon coherent receiver,” IEEE Photonics Technol. Lett. 28(19), 2070–2073 (2016). [CrossRef]
13. H. Zhu, L. Zhou, T. Wang, L. Liu, C. Wong, Y. Zhou, R. Yang, X. Li, and J. Chen, “Optimized silicon QPSK modulator with 64-Gb/s modulation speed,” IEEE Photonics J. 7(3), 1–6 (2015). [CrossRef]
14. P. Dong, C. Xie, L. Chen, L. L. Buhl, and Y. K. Chen, “112-Gb/s monolithic PDM-QPSK modulator in silicon,” Opt. Express 20(26), B624–B629 (2012). [CrossRef] [PubMed]
15. A. Shastri, C. Muzio, M. Webster, G. Jeans, P. Metz, S. Sunder, B. Chattin, B. Dama, and K. Shastri, “Ultra-low-power single-polarization QAM-16 generation without DAC using a CMOS photonics based segmented modulator,” J. Lightwave Technol. 33(6), 1255–1260 (2015). [CrossRef]
16. J. Lin, H. Sepehrian, L. A. Rusch, and W. Shi, “CMOS-compatible silicon photonic IQM for 84 gbaud 16QAM and 70 gbaud 32QAM,” in 2018 Optical Fiber Communications Conference and Exposition (OFC) (Optical Society of America, 2018) pp. 1–3.
17. Z. Wang, Y. Gao, J. Cartledge, E. Huante-Ceron, D. Logan, and A. Knights, “DSP-enabled 104-Gb/s 16-QAM nyquist subcarrier modulation using a silicon micro-ring resonator,” IEEE Photonics Technol. Lett. 30(17), 1571–1574 (2018). [CrossRef]
18. J. Ding, S. Shao, L. Zhang, X. Fu, and L. Yang, “Silicon 16-QAM optical modulator driven by four binary electrical signals,” Opt. Lett. 42(8), 1636–1639 (2017). [CrossRef] [PubMed]
19. D. Perez-Galacho, L. Bramerie, C. Baudot, M. Chaibi, S. Messaoudene, N. Vulliet, L. Vivien, F. Boeuf, C. Peucheret, and D. Marris-Morini, “QPSK modulation in the O-band using a single dual-drive Mach–Zehnder silicon modulator,” J. Lightwave Technol. 36(18), 3935–3940 (2018). [CrossRef]
20. A. Samani, M. Chagnon, D. Patel, V. Veerasubramanian, S. Ghosh, M. Osman, Q. Zhong, and D. Plant, “A low-voltage 35-GHz silicon photonic modulator-enabled 112-Gb/s transmission system,” IEEE Photonics J. 7(3), 1–13 (2015). [CrossRef]
21. A. Samani, D. Patel, M. Chagnon, E. El-Fiky, R. Li, M. Jacques, N. Abadía, V. Veerasubramanian, and D. V. Plant, “Experimental parametric study of 128 Gb/s PAM-4 transmission system using a multi-electrode silicon photonic Mach Zehnder modulator,” Opt. Express 25(12), 13252–13262 (2017). [CrossRef] [PubMed]
22. R. Ding, Y. Liu, Q. Li, Y. Yang, Y. Ma, K. Padmaraju, A. Lim, G. Lo, K. Bergman, T. Baehr-Jones, and M. Hochberg, “Design and characterization of a 30-GHz bandwidth low-power silicon traveling-wave modulator,” Opt. Commun. 321, 124–133 (2014). [CrossRef]
23. D. Patel, V. Veerasubramanian, S. Ghosh, W. Shi, A. Samani, and D. V. Plant, “A 4×4 fully non-blocking switch on SOI based on interferometric thermo-optic phase shifters,” in Optical Interconnects Conference (IEEE, 2014), pp. 104–105. [CrossRef]
24. Y. Zhang, S. Yang, Y. Yang, M. Gould, N. Ophir, A. E. Lim, G. Q. Lo, P. Magill, K. Bergman, T. Baehr-Jones, and M. Hochberg, “A high-responsivity photodetector absent metal-germanium direct contact,” Opt. Express 22(9), 11367–11375 (2014). [CrossRef] [PubMed]
25. M. Jacques, A. Samani, D. Patel, E. El-Fiky, M. Morsy-Osman, T. Hoang, M. G. Saber, L. Xu, J. Sonkoly, M. Ayliffe, and D. V. Plant, “Modulator material impact on chirp, DSP, and performance in coherent digital links: comparison of the lithium niobate, indium phosphide, and silicon platforms,” Opt. Express 26(17), 22471–22490 (2018). [CrossRef] [PubMed]
26. E. El-Fiky, M. Osman, A. Samani, C. Gamache, M. H. Ayliffe, J. Li, M. Jacques, Y. Wang, A. Kumar, and D. V. Plant, “First demonstration of a 400 Gb/s 4λ CWDM TOSA for datacenter optical interconnects,” Opt. Express 26(16), 19742–19749 (2018). [CrossRef] [PubMed]
27. A. Samani, V. Veerasubramanian, E. El-Fiky, D. Patel, and D. Plant, “A silicon photonic PAM-4 modulator based on dual-parallel Mach–Zehnder interferometers,” IEEE Photonics J. 8(1), 1–10 (2016). [CrossRef]
28. P. Cheben, J. H. Schmid, S. Wang, D.-X. Xu, M. Vachon, S. Janz, J. Lapointe, Y. Painchaud, and M.-J. Picard, “Broadband polarization independent nanophotonic coupler for silicon waveguides with ultra-high efficiency,” Opt. Express 23(17), 22553–22563 (2015). [CrossRef] [PubMed]
29. M.-J. Picard, Y. Painchaud, C. Latrasse, C. Larouche, F. Pelletier, and M. Poulin, “Novel spot-size converter for optical fiber to sub-µm silicon waveguide coupling with low loss, low wavelength dependence and high tolerance to alignment,” in Proceedings of European Conference on Optical Communication (ECOC) (IEEE, 2015) pp. 1–3.
