Open Access
Add to BibSonomy
Abstract
We investigated the performance of a silicon photonic device integrating pn-junction-type Mach-Zehnder interferometer optical switches with wavelength-division-demultiplexing and multiplexing (WDM-DEMUX and MUX) filters that perform switching operations to transmit packet data between edge computing. After separating the 1535 and 1565 nm signals with the Mach-Zehnder delay interferometer-type WDM-DEMUX/MUX filters, cascaded dynamic switching operation with a few nanoseconds turn-on/off for On-Off keying (OOK) signals at 10 Gbps were achieved. The power penalty difference between the input and output was within 0.5 dB for 1565-nm signals. We also verified the switching operation at a bit rate of 25 Gbps and observed clear eye openings.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
In the Beyond fifth-generation (5G) world, further development of the internet of things is expected to lead to a significant evolution toward a cyber-physical system (CPS) world, where real space and cyberspace are more closely connected. As the amount of information processed is expected to increase for applications in autonomous driving, remote control, and disaster prediction, the conventional method of sending data to cloud computing (large-scale data center) for processing is inefficient in terms of application requirements. This is because cloud computing is located far from end-users, and a large latency is required for transmission. Therefore, edge computing (micro-data centers (µ-DCs)) should be located on the wireless side of the optical communication path. From this, data processing is expected to reduce latency, improve reliability, and increase bandwidth [1,2]. Recently, from the aspect of improving the data utilization resources, resource disaggregation in several types of computing resources such as, central processing units, graphics processing units, and memories, has been investigated in datacenters [3–6]. The disaggregation concept has also been investigated in metro networks [7–9]. Moreover, resource slicing has been studied in 5G wireless networks to enhance the effect of limited resource utilization in the network [10–13]. We are pursuing the possibility of extending the concept of resource disaggregation and network slicing inside the computing platform to separate resources in µ-DCs connected through optical networks. Figures 1(a) and 1(b) show the network configuration, so called edge/cloud computing platform, and image of an optical node (O-Node) operation, respectively. µ-DCs are connected with multi-core fibers (MCFs) to increase their capacity and eliminate data format conversions from high-speed to low-speed interfaces and vice versa. Research on MCFs has been conducted in the past decade, and it has contributed to an increase in the total capacity of a single fiber owing to space-division multiplexing [14]. In addition, the reduction of light intensity in a fiber can result in avoiding the fiber fuse phenomenon regardless of the increment of the capacity. This is because the signals are distributed in several cores inside the MCF. Recently, the property of large capacity of MCFs has been applied to metro-core networks to increase the total capacity of the network [8,15]. In other applications of MCFs, high-density transceivers in co-packaged optics using vertical-cavity surface-emitting lasers have been demonstrated [16,17]. This application would work to eliminate the data-rate conversion in the time-division/wavelength-division multiplexed signals, and thus the latency for such an operation would be reduced. In our study of the MCF used in the µ-DC connection, the advantage in the elimination of signal processing latency is targeted. In addition, optical switches integrated with wavelength-division-demultiplexing and multiplexing (WDM-DEMUX and MUX) filters are installed between the input/output-side of the MCFs and fan-in/fan-out devices for avoiding the packet collusion between the uplink data from wireless networks and downlink data from the µ-DCs, for example. The concept of the edge/cloud covers a few wireless network areas with multicore fiber connected edge computing through O-Nodes. Thus, uplink and downlink signals have different wavelengths like passive optical network systems. In the future, the number of wavelengths would increase for achieving connection between more number of inputs and outputs. Data from the µ-DC, adjacent O-Node, and wireless network are connected to each destination. The resource capacity and physical size of edge computing are smaller than those of cloud computing, and edge computing has limited processing resources and capabilities. Therefore, an optical switching system should be considered to support high-capacity services by selecting and sending optical packets to multiple edge computers.
Fig. 1. Conceptual image of the edge/cloud computing platform: (a) Network configuration diagram; (b) O-Node structure and operation.
To connect the µ-DCs and use the required resources in each µ-DC in different locations, optical switches are advantageous because of their large bandwidth, format transparency, and high-speed switching operation. Various switching principles have been studied and improved for higher speed, lower power consumption, smaller size, larger capacity, and higher reliability. The switch principles are planar-lightwave-circuit-based Mach-Zehnder interferometers (MZI) multi-stage switches [18–20], micro-electromechanical switches [21,22], semiconductor optical gates or amplifiers (SOAs) [23,24], semiconductor-integrated phased-arrays [25], and silicon photonic switches [26–31]. Among these techniques, silicon photonic integrated devices have several advantages such as high integration and low power consumption, owing to the large refractive index difference and the resultant capability of small bending radius design. The operating principles of optical switches are mainly categorized as follows: thermo-optic effect [26–28], carrier plasma effect with current injection [28–30], and carrier depletion with reverse-biased driving conditions [32]. To achieve a high network utilization efficiency and switch data on a packet-by-packet basis, a nanosecond-order switching operation is required. The application of optical switches in datacenters has been demonstrated owing to their large-capacity interconnection between resources by using SOAs [33], silicon photonics micro-ring resonators [32], and wavelength selective switches [34]. However, few reports have been presented on the clear configurations and operational demonstrations of optical switching systems between edge computing, as shown in Figs. 1(a) and 1(b). From the device point of view, the integrated silicon photonic device consists of a fundamental fabric with several switches and WDM-DEMUX/MUX filters between single-mode fibers (SMFs) and arrayed fabrics existing between MCFs. The structure could be scaled with increasing the number of cores and wavelengths, and this structure is different from the two-dimensional switches containing just the MZIs. We report the fundamental concept of the edge/cloud computing platform and the preliminary results of the silicon photonics optical switches (OSWs) [35]. However, precise evaluation of the performance of the device have not been reported so far.
In this study, silicon photonic switches integrated with WDM-DEMUX/MUX filters are fabricated and evaluated in detail. The performance of the switching operation at both 10 and 25Gbps non-return-to-zero On-Off keying (NRZ-OOK) signals is presented.
The remainder of this study is organized as follows: in Section 2, the device structure is explained; in Section 3, the experimental setup and results are presented (static switching, dynamic switching operation and evaluation of the signal quality are provided); and finally, this study is concluded in Section 4.
2. Device structure
Figure 2(a) shows a silicon photonic integrated device for the optical switching system. A photograph of this device is shown in Fig. 2(b). The waveguide consists of a 0.44 µm wide and 0.22 µm thick silicon core region surrounded by upper and lower SiO2 cladding regions. The device size is 2 mm × 1.5 mm with three input ports and four output ports. The switching element consists of three Mach-Zehnder interferometer optical switches (MZI-OSWs) with pn junctions and rib structures. Mach-Zehnder delay interferometer (MZDI)-type WDM-DEMUX/MUX filters are connected at the input and output sides of one of the OSWs. All splitters and couplers used in MZI and MZDI are multi-mode-interferometers. In the MZI-OSW, the phase-shifter is 0.5 mm long. The carrier plasma effect [29–31] is used for the phase shifting for switching. We aim to realize packet-by-packet switching between each destination for packets with an Ethernet frame; therefore, switching the transient time by a few nanoseconds is required for a bit rate of 25 Gbps and higher. Accordingly, we selected the pn-junction-type MZI structure for the silicon photonic platform. For the pn-junction-type silicon photonic switches, both forward- and reverse-biased operations have been reported [32]. The reverse-biased mode can realize faster operation above 25 Gbps; however, the driving voltage should typically exceed 10 V in a device with short-length phase shifters. Nevertheless, a forward-biased operation requires a larger driving current, but the efficiency of the phase shift is greater than that of the reverse bias case and a shorter phase-shifter is available. Accordingly, we selected forward-biased conditions. A high integration can be attained because of the small-sized phase shifters. In the WDM-DEMUX/MUX filter, we selected the MZDI because the structure is simple and small, we just design the delay length according to a free spectral range (FSR), and tolerance against temperature change can be obtained due to sinusoidal transmissivity characteristics. Delay length of the WDM-DEMUX/MUX filter is designed for splitting/coupling signals at wavelengths of 1565 and 1535 nm, with a FSR of 60 nm. We selected both wavelengths because they are the shortest and the longest wavelength in the C-band, and thus the tolerance of the filter performance against fabrication error and temperature change could be large. We designed the device to operate at a wavelength of 1550 nm-range instead of 1310 nm-range because of our facility environment. The wavelength would be changed to the 1310 nm band owing to low chromatic dispersion in the future. We believe that switching, filtering, and design concepts do not change. The thermo-optic effect is used for the phase shift. The delay length $\Delta L$ is designed according to the following equation:
where $\mathrm{\lambda }$ is the wavelength, ${n_g}$ is the group velocity of the waveguide, and $\Delta \mathrm{\lambda }$ is the filter FSR. In this experiment, an FSR of 60 nm (operating wavelengths of 1535 and 1565 nm) was set; thus, a $\Delta L$ of 9.5 µm was designed.Fig. 2. Schematic of the fabricated device: (a) three optical switches and WDM-DEMUX/MUX filters integrated in one chip; (b) photograph of the device.
As shown in Fig. 2(a), signals from a µ-DC at a wavelength of 1565 nm are injected into input port #1, and they are switched with OSW1 to the wireless network (output port #1) or to one of the input ports of OSW2. The signals are then switched to the µ-DC (output port #2) for loopback or to the output-side adjacent O-Node, followed by the WDM-MUX filter (output port #3). Signals from the input-side adjacent O-Node are multiplexed signals at 1565 and 1535 nm wavelengths. They are split into signals at 1565 and 1535 nm, and they go to the upper and lower output of the WDM-DEMUX filter. The former is switched from OSW2 to output ports #2 or #3 through the WDM-MUX filter. The latter is switched from OSW3 to output port #3 through the WDM-MUX filter or port #4. Signals from the wireless networks at a wavelength of 1535 nm are injected into input port #3. The signal is then switched from OSW3 to the output, followed by a WDM-MUX (output port #3) or to the µ-DC (output port #4).
3. Experiments
3.1 Performance evaluation for continuous wave (CW) input light
The relationship between transmittance and applied voltage was measured as the extinction ratio and crosstalk of the fabricated integrated devices, and the optimum operating voltage for the WDM-DEMUX/MUX filters was determined. In this study, only one-sided phase modulation was applied. Measurements were made with a forward-biased voltage applied to obtain a sufficient phase change.
The experimental setup is shown in Fig. 3. CW light was emitted from a laser source, and transverse electric (TE) polarization was input to the device through a lensed fiber controlled with a polarizer and polarization controller (PC). The output power from the device was measured with an optical power meter (OPM) while varying the direct current (DC) voltage applied to the optical switch and WDM-DEMUX/MUX filters. As the operating mode of the optical switch was forward bias to the pn junction, the output power versus the injection current was drawn for characterization.
Fig. 3. Experimental setup of input/output characteristics with DC voltage applied to O-SWs under CW condition. PC: Polarization controller: optical power meter (OPM).
First, the switching characteristics of the optical switches were measured. Near-field patterns of the light injected into input port #1 were tested, and the outputs from ports #1, #2, and #3 were observed to obtain the maximum intensity in each output port according to the current injected into OSW1 and OSW2, as shown in Fig. 4. Switching can be observed between output ports #1, #2, and #3 with a certain extinction ratio. The output light power as a function of the current injected into the optical switches is shown in Fig. 5. It is noted that each OSW has integrated termination resister at the end of the MZI structure in parallel, and the series resistance was evaluated to be 109 Ω. On the other hand, the series resistance of the pn-junction of the phase-shifter was 54 Ω. We designed the termination resister for reverse-biased driving condition, but the forward-biased condition was selected. Therefore, current through the termination resister could be eliminated, and the electrical power consumption could be reduced in the future design. The optical output before the fiber on the input side of the silicon photonic device was 6.8 dBm. As shown in Fig. 5(a), 1565 nm wavelength light was injected into input port #1, and OSW1 was switched to one of the output ports and to the other side with the injection current into OSW1. In the case of switching from the output connected with the OSW2 input, light was switched from outputs #2 and #3 with the injection current into OSW2. From these results, the characterization is summarized in Table 1. The extinction ratios, crosstalks, and losses (including coupling loss at both facets) were 16.9, 31.5, and 9.4 dB for output 1; 14.5, 18.8, and 12.8 dB for output 2; and 18.6, 13.1 and 13.2 dB for output 3, respectively. Similar to Fig. 5(a), the output power as a function of the injection current for OSW2 for an input light from input port #2 at a wavelength of 1565 nm, for OSW3 for an input light from input port #2 at a wavelength of 1535 nm, and for an input light from input port #3 at a wavelength of 1535 nm is shown in Figs. 5(b)–(d). In addition, the extinction ratio, crosstalk, and insertion loss in each operation are summarized in Tables 1–4. The extinction ratios were from 12.2 to 19.3 dB, crosstalks from 11.5 to 31.5 dB, and losses from 9.4 to 13.2 dB. In our proposed edge/cloud computing network, we consider that two O-Nodes are connected typically, and optical signals transmit four OSWs in maximum.
Fig. 4. Near-field pattern of light at a wavelength of 1565 nm injected into input port #1: (a) output from port #1; (b) output from port #2; (c) output from port #3.
Fig. 5. Output power as a function of injection current to the optical switch: (a) input light at a wavelength of 1565 nm into port #1; (b) input light at a wavelength of 1565 nm into port #2; (c) input light at a wavelength of 1535 nm into port #2; (d) input light at a wavelength of 1535 nm into port #3.
Table 1. Summary of the characterization of optical switches: input light at a wavelength of 1565 nm into port #1
Table 2. Summary of the characterization of optical switches: input light at a wavelength of 1565 nm into port #2
Table 3. Summary of the characterization of optical switches: input light at a wavelength of 1535 nm into port #2
Table 4. Summary of the characterization of optical switches: input light at a wavelength of 1535 nm into port #3
Under this condition, crosstalk of the OSW of more than 20dB at each operating wavelength would be required. The deviation of each characteristic originated from the free carrier loss in the current-injected region, as well as the non-ideal splitting ratio of the input and output splitters. The estimated carrier density into the pn-junction is about the order of 1018 cm-3, the transmissivity through 500 µm-long phase-shifter is estimated to be about 82% by using Drude model [36]. In addition, splitting ratio of the 2 × 2 MMI was evaluated to be 45.8:54.2 and 52.4:47.6 (bar:cross)at wavelengths of 1535 nm and 1565 nm, respectively. From these situations, the worst extinction ratio would be degraded to be 18 dB. If the splitting ratio is intentionally asymmetric to compensate the absorption loss in the phase-shifter, the extinction ratio without current injection would be 21 dB. We should care much about the balance of the extinction ratio between on and off state, but asymmetric splitting ratio would be one of the effective methods for improving the device performance. We could improve them by redesigning the length of the phase-shifted region and splitters. As for the optical loss, minimum received sensitivity at the receiver is -19 dBm in average without optical amplifiers, and the power budget is about 22 dB in case of the optical launches power of +3 dBm. In this case, optical loss of each OSW should be less than 11 dB for two cascaded MZIs including the coupling losses for both facets. Therefore, reduction of optical loss should be needed, and the decrease of the coupling loss is one of the possibilities.
Next, we evaluated the performance of MZDI-type WDM-DEMUX/MUX filters. Figure 6 shows the transmissivity as a function of wavelength under the heater’s applied voltage of 0 and 3.7 V. Without the applied voltage, the maximum and minimum transmissivity for the upper and lower outputs were approximately 1550 nm, which deviated from the designed wavelengths of 1535 and 1565 nm. By applying a voltage of 3.7 V, 1535, and 1565 nm, wavelengths were output from the lower and upper output with extinction ratios of 12 and 20 dB, respectively. In our proposed edge/cloud computing network, on/off ratio of the WDM-DEMUX filter of more than 20dB at each operating wavelength would be required, assuming that optical amplifiers are installed after each O-Node for compensating loss of the optical switch. To improve the performance, redesign of the FSR of the filter would be effective because the present FSR is larger than that of the desired one. In addition, the transmissivity of the WDM-DEMUX and MUX filters with the applied voltage is shown in Figs. 7(a) and 7(b), respectively. The optimum operating voltage of the WDM-DEMUX/MUX filter was determined from the voltage values that maximize the transmission of two wavelengths. It was approximately 3.2 V for the DEMUX (output port #4 at 1535 nm and #2 at 1565 nm for input port #2 injection) and 0 or 5V for the MUX filter (output port #3 for input port #3 at 1535 nm and for input port #2 at 1565 nm). The series resistance of the heater for the WDM-DEMUX-MUX was 670 Ω, and thus the electrical consumption power was 15 mW and 37 mW at the applied voltage of 3.2 V and 5.0 V, respectively. As can be seen in Fig. 7(b), the voltage condition could be close to 0V. In addition, the operating voltage for WDM-DEMUX filter indicated in Fig. 7(a) could be flipped by changing the delay line to the opposite side in the MZDI. Then, the reduction of electrical consumption power could be expected.
3.2 Performance evaluation for radio frequency (RF) signal application
Next, we measured the output waveform of the RF signal applied to the 10 Gbps 27– 1 pseudo-random bit sequence (PRBS) NRZ-OOK signal input. Figure 8 shows the experimental setup. The 10 Gbps electrical signal generated by the pulse pattern generator (PPG) was input into the lithium niobite modulator, which modulated the CW output light from the wavelength-tunable laser and generated a 10 Gbps NRZ-OOK optical signal. The signal was adjusted to TE polarization by using a polarizer and a PC. The signal passed between each input and output of the optical switch element, which was amplified by an EDFA. Excessive amplified spontaneous emissions were eliminated with a BPF, and they were input into a photodetector (PD). The PD output was input to the sampling oscilloscope to observe the time waveform.
Fig. 8. Experimental setup of the time waveform for dynamic operation. PPG: Pulse Pattern Generator, EDFA: Erbium Doped Fiber Amplifier, BPF: Band Pass Filter, VOATT: Variable Optical Attenuator, PD: Photodiode, PC: Polarization Controller, OSC: Oscilloscope.
Then, we measured the dynamic switching behavior of a 1565 nm signal injected into input port #1. The yellow and green curves depicted in Fig. 9(a) represent the signals applied to OSW1 and OSW2, with amplitudes of 0.15 and 0.18 V, respectively. The driving conditions of OSW1 and 2 are summarized in Table 5. The series resistance of each MZI was 54 Ω, and the resister of the terminator was 109 Ω. And thus, the average electrical consumption power were 14 mW and 16 mW for OSW1 and OSW2, respectively, considering the 1.0 V built-in voltage of Si pn-junction. The electrical consumption power in the pn-junction were 0.58 mW and 1.5 mW for OSW1 and OSW2, respectively, and therefore power reduction could be expected by eliminating the termination resistance because nanosecond-order operation is sufficient. The pulse widths and periods were 100 and 200 ns, respectively, for both signals. The timing of switching between the higher and lower voltage was 50 ns, which was intentionally shifted for the voltage applied to OSW2 as compared with that of OSW1. The output time waveform was observed using an oscilloscope. As shown in Figs. 9(b)–(d) for outputs #1, #2, and #3, the output waveform corresponding to the applied RF pulse was successfully achieved. Compared between the output power of Figs. 9(b) and 9(c), the loss of the MZI-OSW was estimated to be 3.4 dB. As shown in the inset of Fig. 9(b), clear eye openings were observed. The rise and fall times were close, as indicated in Figs. 10(a) and 10(b), and we confirmed that the rise and fall times were 6 ns and 4.5 ns, respectively. We also examined the operation of signals at a bit rate of 25 Gbps. Switching operation and clear eye openings were achieved, as indicated in Figs. 11(a) and 11(b).
Fig. 9. Time waveform of dynamic operation of optical switches: (a) electrical pulses; (b) output #1 from OSW1; (c) output #2 from OSW2; (d) output #3 from OSW3.
Fig. 11. Time waveform of optical switch at a bit rate of 25 Gbps:(a) switched signal; (b) magnified waveform.
Table 5. Operating conditions of OSW1 and OSW2
3.3 BER measurement for pseudo-random codes
Subsequently, the BER performance of the output light through the integrated device was measured. The experimental setup is shown in Fig. 12. The BER evaluation of signals transmitted between each input and output of the optical switch element for 10 Gbps 27–1 PRBS NRZ-OOK signals was examined under DC current injected into the OSWs.
The measured results of the BER of the signal through the optical switch are shown in Fig. 13 for the 1565 nm signal injected into inputs #1 and # 2. The back-to-back signal was also measured for comparison. All the output signals from output ports 1 to 3 were observed at a BER of less than 10−9, and the power penalty of each case was less than 0.7 dB. Table 6 summarizes the power penalty for each output of the 1565 nm wavelength input. The average power penalty of 0.52 dB was verified.
Fig. 13. BER measurement results (at a bit rate of 10 Gbps for a 1565 nm input signals): (a) OSW1 for input 1 signal; (b) OSW1 for input 2 signal.
Table 6. Summary of power penalty for 1565 nm wavelength input signals at a bit rate of 10 Gbps
4. Conclusion
We fabricated and characterized a silicon photonic device using a pn-junction-type MZI optical switch integrated with MZDI-type WDM-DEMUX/MUX filters. We selected a forward-biased operation using short-length phase shifters and lower driving voltages. The extinction ratios were from 12.2 to 19.3 dB, crosstalks from 11.5 to 31.5 dB, and losses from 9.4 to 13.2 dB, respectively, for each path. The extinction ratio and crosstalk are expected to improve by lowering the operating voltage and reducing carrier absorption.
In addition, switching operations for NRZ-OOK signals at bit rates of 10 and 25 Gbps, and switching turn-on/off times of a few nanoseconds were observed. A faster switching operation can be expected by applying a voltage to both electrodes in a push-pull mode on each optical switch. For the 10 Gbps signals, a BER of less than 10−9 was achieved between all inputs and outputs for 1565 nm signals. From these results, we confirmed the possibility of using a high-speed, low-power-penalty optical switching device for next-generation edge-computing platforms.
Funding
National Institute of Information and Communications Technology (00101).
Acknowledgments
We would like to acknowledge Prof. Emeritus K. Iga, Prof. K. Kobayashi, Prof. F. Koyama, and Assoc. Prof. T. Miyamoto for their encouragement and discussions. This work has been supported by Research fund of “Beyond 5G research development promotion project #00101” from NICT.
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. B.P. Rimal, D.P. Van, and M. Maier, “Mobile edge computing empowered fiber-wireless access networks in the 5G era,” IEEE Commun. Mag. 55(2), 192–200 (2017). [CrossRef]
2. B.P. Rimal, D.P. Van, and M. Maier, “Mobile-edge computing versus centralized cloud computing over a converfged FiWi access network,” IEEE Trans. Network and Service Management 14(3), 498–513 (2017). [CrossRef]
3. G. Zervas, H. Yuan, A. Saljoghei, Q. Chen, and V. Mishra, “Optically disaggregated data centers with Minimal remote memory latency: Technologies, architectures, and resource allocation,” J. Opt. Commun. Netw. 10(2), A270–A285 (2018). [CrossRef]
4. R. Lin, Y. Cheng, M. D. Andrade, L. Wosinska, and J. Chen, “Disaggregated data centers: Challenges and Trade-offs,” IEEE Commun. Mag. 58(2), 20–26 (2020). [CrossRef]
5. R. Takano and K. Suzaki, “Disaggregated acceleratoer management system for cloud data centers,” IEICE trans. Inf. & Syst. E104.D(3), 465–468 (2021). [CrossRef]
6. M. Fariborz, X. Xiao, P. Fotouhi, R. Proietti, and S.J.B. Yoo, “Silicon photonic Flex-LIONS for reconfigurable multi-GPU systems,” J. Lightwave Technol. 39(4), 1212–1220 (2021). [CrossRef]
7. K. Ishii and S. Namiki, “mathematical model of optical functional blocks for automating fully disaggregated optical networks,” Proc. Optical Fiber Communication Conf 2021 (OFC2021), F1C.1, San Diego (2021).
8. C. Manso, R. Muñoz, N. Yoshikane, R. Casellas, R. Vilalta, R. Martínez, R. Tsuritani, and I. Morita, “TAPI-enabled SDN control for partially disaggregated multi-domain (OLS) and multi-layer (WDM over SDM) optical networks,” J. Optical. Commun. And Netwirking 13(1), A21–A33 (2021). [CrossRef]
9. M. Kawashima, “IOWN for digital twin enabled societies,” Proc. Optical Fiber Communication Conf 2022 (OFC2022), Tu3G.5, San Diego (2022).
10. C.-Y. Chang, N. Nikaein, O. Arouk, K. Katsalis, A. Ksentini, T. Turletti, and K. Samdanis, “Slice orchestration for multi-service disaggregated ultra-dense RANs,” IEEE Comm. Mag. 56(8), 70–77 (2018). [CrossRef]
11. B. Shariati, J.J. Pedreno-Manresa, A. Dochhan, A.S. Muqaddas, R. Casellas, O. González de Dios, L.L. Canto, B. Lent, J.E. López de Vergara, S. López-Buedo, F.J. Moreno, P. Pavón, L. Valasco, S. Patri, A. Giorgetti, F. Cugini, A. Sgambelluri, R. Nejabati, D. Simeonidou, R.-P. Braun, A. Autenrieth, J.P. Elbers, J.K. Fischer, and R. Freund, “A latency-aware real-time video surveillance demo: Network slicing for improving public safety,” Proc. Optical Fiber Communicatoin 2021 (OFC2021), F4H.2, San Diego (2021).
12. D. Larrabeiti, G. Otero, J.P. Fernández-Palaclos, L.M. Contreras, and J.A. Hernández, “Latency-aware network, architecture for 5G backhaul and fronthaul,” Proc. Optical Fiber Communicatoin 2021 (OFC2022), Tu2G.3, San Diego (2022).
13. C. Wang, X. Xue, N. Yoshikane, F. Balasis, H. Guo, and T. Tsuritani, “Fiber-to-application: Optical slicing to enhance application performance over a metro transport network,” Proc. Optical Fiber Communicatoin 2021 (OFC2022), W4F.5.3, San Diego (2022).
14. B. J. Puttnam, G. Rademacher, R.S. Luís, T.A. Eriksson, W. Klaus, Y. Awaji, N. Wada, K. Maeda, S. Takasaka, and R. Sugizaki, “High data-rate and long distance MCF transmission with 19-core C + L band cladding-pumped EDFA,” J. Lightwave Technol. 38(1), 123–130 (2020). [CrossRef]
15. M. Jinno, T. Kodama, and T. Ishikawa, “Principle, design, and prototyping of core selective switch using free-space optics for spatial channel network,” J. Lightwave Technol. 38(18), 4895–4905 (2020). [CrossRef]
16. P. Maniotis, L. Schares, D.M. Kuchta, and B. Karacali, “Improving data center network locality w/ co-packaged optics,” Proc. 47th European Conference on Optical Communications (ECOC2021), We1F.1, Bordeaux (2021).
17. L. Dong, X. Gu, and F. Koyama, “Thermal Crosstalk Evaluation of 1.1 µm-band Vertical Cavity Surface Emitting Laser Array for Multi-core Fiber Transmission,” Proc. Microoptics Conference 2021, PO-34 (online) (MOC2021).
18. K. Suzuki, T. Yamada, O. Moriwaki, H. Takahashi, and M. Okuno, “Polarization-Insensitive Operation of Lithium Niobate Mach–Zehnder Interferometer With Silica PLC-Based Polarization Diversity Circuit,” IEEE Photonics Technology Letters 20(10), 773–775 (2008). [CrossRef]
19. T. Goh, M. Yasu, K. Hattori, A. HImeno, M. Okuno, and Y. Ohmori, “Low-loss and high extinction ratio strictly nonblocking 16×16 thermooptic matrix switch on 6-in wafer using silica-based planar lightwave circuit technology,” J. Lightwave Technol. 19(3), 371–379 (2001). [CrossRef]
20. K. Sato, “realization and application of large-scale fast optical circuit switch for data center networking,” J. Lightwave Technol. 36(7), 1411–1419 (2018). [CrossRef]
21. J. Kim, C.J. nuzman, B. Kumar, D.F. Lieuwen, J.S. Kraus, A. Weiss, C.P. Litchtenwalner, AS.R. Papazian, R.E. Frahm, N.R. Basavanhally, D.A. Ramsey, V.A. Aksyuk, F. Pardo, M.E. Simon, V. Lifton, H.B. Chan, M. Haueis, A. Gasparyan, H.R. Shea, S. Arney, C.A. Bolle, P.R. Kolodner, R. Ryf, D.T. Neilson, and J.V. Gates, “1100×1100 port MEMS-based optical crossconnect with 4-dB maximum loss,” IEEE Photonics Technol. Lett. 15(11), 1537–1539 (2003). [CrossRef]
22. N. Quack, T.J. Seok, S. Han, R.S. Muller, and M.C. Wu, “Scalable row/column addressing of silicon photonic MEMS switches,” IEEE Photonics Technol. Lett. 28(5), 561–564 (2016). [CrossRef]
23. M. Zegaoui, N. Choueib, J. Harari, D. Decoster, V. Magnin, X. Wallart, and J. Chazelas, “2×2 InP Optical Switching matrix based on carrier-induces effects for 1.55-mm applications,” IEEE Photon. Techlogy Lett. 21(19), 1357–1359 (2009). [CrossRef]
24. Q. Cheng, A. Wonfor, J. L. Wei, R. V. Penty, and I. H. White, “Low-energy, high-performance lossless 8×8 SOA switch,” 2015 Optical Fiber Communications Conference and Exhibition (OFC), 2015, pp. 1–3.
25. I. M. Soganci, T. Tanemura, K. Takeda, M. Zaitsu, M. Takenaka, and Y. Nakano, “Monolithic InP 100-port photonic switch,” 36th European Conference and Exhibition on Optical Communication, 2010, pp. 1–3.
26. T. Fujisawa, J. Takano, Y. Sawada, and K. Saitoh, “Low-loss and small 2×4λ multiplexers based on 2×2 and 2×1 Mach-Zehnder interferometers with on-chip polarization multiplexing for 400GbE,” J. Lightwave Technol. 39(1), 193–200 (2021). [CrossRef]
27. K. Suzuki, R. Konoike, S. Suda, H. Matsuura, S. Namiki, H. Kawashima, and K. Ikeda, “Low-loss, low-crosstalk, and large-scale optical switch based on silicon photonics,” J. Lightwave Technol. 38(2), 233–239 (2020). [CrossRef]
28. K. Suzuki, R. Konoike, N. Yokoyama, M. Seki, M. Ohtsuka, S. Saitoh, S. Suda, H. Matsuura, K. Yamada, S. Namiki, H. Kawashima, and K. Ikeda, “Nonduplicate polarization-diversity 32×32 silicon photonics switch basd on a SiN/Si double-layer platform,” J. Lightwave Technol. 38(2), 226–232 (2020). [CrossRef]
29. D. Pérez-Galacho, D. Marris-Morini, R. Stoffer, E. Cassan, C. Baudot, T. Korthorst, F. Boeuf, and L. Vivien, “Simplified modeling and optimization of silicon modulators based on free-carrier plasma dispersion effect,” Opt. Express 24(23), 26332–26337 (2016). [CrossRef]
30. C.J. Krückel, H. Becker, Y. Ban, M.J.R. heck, J.V. Campenhout, and D.V. Thourhout, “Towards maximum energy efficiency of carrier-injection-based silicon photonics,” J. Lightwave Technol. 39(9), 2931–2940 (2021). [CrossRef]
31. R. Konoike, K. Suzuki, and K. Ikeda, “Path independent insertion-loss (PILOSS) 8×8 silicon photonics swithch with < 8 nsec switching time “, Proc. Optical Fiber Communication Conf 2022 (OFC2022), W4B.3, San Diego (2022).
32. A. Biberman, H.L.R. Lira, K. Padmaraju, N. Ophir, J. Chan, M. Lipson, and K. Bergman, “Broadband silicon photonic electrooptic switch for photonic interconnection networks,” IEEE Photonics Technol. Lett. 23(8), 504–506 (2011). [CrossRef]
33. X. Xue, K. Priftu, B. Pan, S. Chen, X. Guo, F. Yan, S. Zhang, C. Xie, and N. Calabretta, “Experimental assessments of a flexible optical switch and control system with dynamic bandwidth allocation,” Proc. 47th European Conference on Optical Communications (ECOC2021), Tu1F.2, Bordeaux (2021).
34. E.D. Chansky, V. Arrunategul-Norvick, T. Hirokawa, L. Alberto Campos, H. Zhang, M. Xu, Z. Jia, and C.L. Schow, “Edge wavelength selective switch for optical access networks,” Proc. Optical Fiber Communication Conf 2022 (OFC2022), W4B.2, San Diego (2022).
35. N. Matsunaga and H. Uenohara, “Silicon Photonics Optical Switch Performance Integrated with WDM filters for Low-Latency Edge-Computing Platform,” to be presented on International Conference on Photonics in Switching and Computing 2022 (PSC2022), MD3-4, Toyama (2022).
36. R.A. Soref and B.R. Bennett, “Electrooptical effects in silicon,” IEEE J. Quantum Electron.23(1), 123–129 (1987). [CrossRef]
