We fabricated and characterized a silicon photonics 8 × 8 strictly non-blocking optical switch based on double-Mach–Zehnder (MZ) element switches. The double-MZ switches, each of which consisted of an intersection and two asymmetric MZ switches, enabled the suppression of crosstalk across a wide wavelength range. The 8 × 8 switch exhibited an average fiber-to-fiber insertion loss of 11.2 dB and -20 dB crosstalk in a bandwidth wider than 30 nm. Furthermore, we constructed an 8 × 8 polarization-diversity switch by using two 8 × 8 switches and demonstrated 32-Gbaud dual-polarization, quadrature-phase-shift-keying, four-channel wavelength-division-multiplexed signal transmission without significant signal degradation.
© 2017 Optical Society of America
Demands for high-capacity optical communication have been increasing rapidly due to the growing prevalence of video content on the internet. In addition, electric power consumption is becoming a limiting factor for transmission capacity. Thus, we previously proposed a dynamic optical path network (DOPN) as a potential energy-efficient, high-capacity network system and demonstrated its basic operation in a testbed network . The DOPN is based on hierarchal dynamic switching of the optical fiber path, wavelength, and optical data unit. It is desirable that the optical fiber path switch be strictly non-blocking to enable flexible operation. Here, strictly non-blocking means that any input port can be connected with any output port arbitrarily and that an unused input port can always be connected to an unused output port without re-arranging the existing connections. Silicon photonics platforms are among the most suitable platforms for optical path switches because they can offer dense integration and high homogeneity, enabling large-scale and low-cost production.
Several strictly non-blocking Si switches based on different topologies, including the path-independent insertion loss (PILOSS) , switch and select , phased array , and MEMS-actuated cross-point  topologies, have been demonstrated. Among them, we believe that the PILOSS topology remains promising for the realization of switches with larger port counts in terms of insertion loss, power consumption, and long-term stability. So far, we have demonstrated 8 × 8  and 32 × 32 PILOSS switches [7, 8]. In our devices, the bandwidth was not sufficiently broad (~7.5 nm) to achieve crosstalk lower than −20 dB due to the wavelength dependence of the 3 dB directional coupler of the Mach–Zehnder (MZ)-element switches. A broad low-crosstalk bandwidth is important when processing wavelength-multiplexed signals. A straightforward method of achieving broadband operation involves replacing the directional coupler with a broadband coupler such as a multi-mode interference (MMI) coupler, an adiabatic coupler, or a wavelength-independent coupler (WINC). However, we believe that these broadband couplers are not suitable for high-port-count switches because the insertion losses of MMI couplers are too high (typically 0.5 dB), the footprints of adiabatic couplers are not compact (sub-millimeter length), and the fabrication tolerances of WINCs are too severe (<0.1 nm). Thus, we instead modified the switch structure from the conventional MZ switch illustrated in Fig. 1(a) to the double-MZ switch depicted in Fig. 1(b), which was proposed by Goh et al.  in the silica platform.
In this paper, we report on the fabrication and characterization of an silicon photonics 8 × 8 PILOSS switch based on double-MZ switches . The 8 × 8 switch exhibited a fiber-to-fiber insertion loss of 11.2 dB and a bandwidth for −20 dB crosstalk that was wider than 30 nm. In addition, we produced an 8 × 8 polarization-diversity switch and demonstrated 32-Gbaud dual-polarization (DP), quadrature-phase-shift-keying (QPSK), wavelength-division multiplexed (WDM) signal transmission without significant signal degradation.
2. Fabricated switch
Each double-MZ switch consisted of an intersection and two MZ switches, as illustrated in Fig. 1(b). The intersection was formed by a 100% coupling directional coupler in an upright orientation , and the MZ switch had thermooptic phase shifters based on a TiN heater. The MZ switch was asymmetric so that the bar state could be produced without heating. The 3 dB coupler of the MZ switch was a directional coupler. The devices were designed to be used in a transverse-magnetic (TM)-like mode. When the double-MZ switch is in the cross state (connections: port 1–port 2’ and port 2–port 1’), the two MZ switches are in the bar state. Conversely, when the double-MZ switch is in the bar state (connection: port 2–port 2’), the two MZ switches are in the cross state. The port 1–port 1’ connection is not used for the PILOSS configuration depicted in Fig. 2 . In the PILOSS configuration, the double-MZ switches were alternately arrayed in the normal and upside-down orientations.
Figure 3(a) depicts the 8 × 8 switch chip fabricated on a 300-mm-diameter silicon-on-insulator wafer by using a complementary metal–oxide–semiconductor process line equipped with an immersion ArF scanner. The Si waveguides were 0.43 μm wide, 0.22 μm thick, and buried underneath 1.5-μm-thick SiO2 over-cladding. The TiN heaters, which served as thermooptic phase-shifters, were fabricated on the over-cladding and buried underneath a 0.2-μm-thick SiO2 insulating layer. The 1.5-μm separation between Si waveguides and TiN heaters was confirmed to be enough to avoid the absorption by the metal, using simulations and experiments. Al-Cu electric wirings were patterned on the insulating layer and were passivated with a 0.6-μm-thick SiOx passivation layer. The SiOx layer was removed from above the electrode pads to form electrical contacts. The footprint of each 8 × 8 double-MZ switch matrix was 1.7 mm × 7.7 mm.
The switch chip was divided into 13 mm × 13 mm sections and then die-bonded onto a ceramic chip carrier, as illustrated in Fig. 3(b). The electrode pads on the chip were wire-bonded to the pads on the chip carrier. Edge couplers and an optical fiber array were used for optical coupling. The edge couplers had the standard inverse-taper shape (width: 0.15 μm, length: 100 μm). The optical fiber array consisted of 20 high-delta fibers with mode-field diameters of 4.5 μm. The coupling loss between each edge coupler and high-delta fiber was estimated to be 2.5 dB/facet. The chip carrier, which had a 304 pin grid array (PGA) on the back side, was inserted into a PGA socket on a printed circuit board (PCB) with control electronics, as shown in Fig. 3(c). All of the MZ switches were operated through a field-programmable gate array (FPGA) and buffer integrated circuits (ICs) on the PCB and calibrated using the method of Suda et al. . The total electric power consumption was approximately 0.8 W, including the switching power (2 × 8 × 40 mW = 0.64 W) and the heating power required for compensation of the initial phase error between the two arms of the MZ switches (0.16 W). This power consumption is less than 1/20 of that previously reported for an 8 × 8 PILOSS switch based on a silica planar lightwave circuit .
3. Results and discussion
3.1 Double-MZ-element switch
Figure 4 presents the measured transmission spectra of the double-MZ-element switch in the cross and bar states. The transmission spectra were measured using a broadband light source and an optical spectrum analyzer. In the cross state shown in Fig. 4(a), the bandwidth for −30 dB crosstalk is wider than 40 nm for both paths. This crosstalk was limited by the intersection, whose coupling ratio was not 100% at any wavelength apart from its operating wavelength. The propagation route of the leakage from port 1 to port 1’ is as follows: port 1 → intersection → front MZ switch → port 2 (reflection at the input facet) → front MZ switch → intersection → port 1’. In the same manner, the propagation route of the leakage from port 2 to port 2’ is as follows: port 2 → front MZ switch → intersection → port 1 (reflection at the input facet) → intersection → rear MZ switch → port 2’. We believe that the crosstalk can be improved by replacing the directional-coupler-based intersection with a broadband intersection . In the bar state depicted in Fig. 4(b), there are two wavelength-dependent leakage paths (paths 1–2’ and 2–1’). However, these leakage paths do not reach any output ports of the PILOSS switch fabric, but rather are terminated at the idle output ports. The electric power required to switch from the bar state to the cross state (π phase shift) was ≈40 mW, and the switching time (10%–90%) was ≈30 μs. The switching power can be reduced to ≈20 mW by changing the TM design to TE design because of the stronger confinement to silicon.
3.2 Fiber-to-fiber insertion losses of all 64 paths
The fiber-to-fiber insertion losses of the 8 × 8 double-MZ switch, which had 82 = 64 paths, were evaluated by using a tunable laser diode, a polarization controller, and an eight-channel optical power meter. The wavelength of the tunable laser diode was set to 1.535 μm, at which the crosstalk of the 8 × 8 switch is minimized, as described in the next section. The polarization of the input light was set to be TM-like by using the polarization controller.
Figure 5 illustrates the measured fiber-to-fiber insertion losses of all 64 paths. The average insertion loss was found to be 11.2 dB, and the on-chip loss, defined as the insertion loss without the fiber-to-chip coupling loss, was estimated to be 6.2 dB. The on-chip loss consisted of the following losses: 2 × 1.4 dB propagation loss of the routing waveguides connecting the edge couplers to the 8 × 8 switch fabric, 0.2 dB insertion loss of the polarization cleaner (100% coupling directional coupler designed for the TM-like mode) placed after the edge couplers, and 3.2 dB insertion loss of the 8 × 8 switch fabric. A single path includes nine MZ switches and 14 intersections. Therefore, the losses of the MZ switch and intersection were estimated to be 0.2 dB and 0.1 dB, respectively. Due to the PILOSS topology, the insertion loss variation was only ± 0.85 dB. We believe that the residual variation originated from fluctuations in the fiber-to-chip coupling loss ( ± 0.65 dB) and the difference between the propagation lengths of the routing waveguides ( ± 0.2 dB). The wavelength dependence in the C-band is less than 3 dB.
3.3 Crosstalk in the worst-case scenario
It is considered that the main contribution to the crosstalk is the leakage at the intersections between paths, and the number of intersections with other paths is related to the crosstalk. Figure 6(a) depicts one of the worst-case paths, that from port 8 to port 7’, which has 11 intersections with other paths. The crosstalk of path 8–7’ was measured using the tunable laser diode, polarization controller, and eight-channel optical power meter. The wavelength of the launched light was changed from 1527 nm to 1565 nm while optimizing the polarization.
The measurement procedure was as follows. First, all of the element switches were set to the states shown in Fig. 6(a). Then, cw light was launched from port 8, and the power output from port 7’ was measured. Next, cw light was launched from port 1, and its leakage to port 7’ was measured. In the same manner, the leakages from input ports 2–7 to output port 7’ were measured. Finally, the crosstalk was estimated as the sum of the leakages from the other paths divided by the power output from path 8–7’.
Figure 6(b) presents the wavelength characteristics of the crosstalk of path 8–7’. The bandwidth for −20 dB crosstalk is wider than 30 nm, i.e., more than four times wider than that previously reported for a single-MZ-based 8 × 8 PILOSS switch . Additionally, the bandwidth for −30 dB crosstalk is 17 nm. The wavelength corresponding to the minimum crosstalk is blue-shifted from the center of the C-band due to the discrepancy between the designed and fabricated switches.
We consider that the wavelength characteristics of the crosstalk originates from the accumulation of the leakages at the directional-coupler intersections, as depicted in Fig. 7, since the leakages from MZ switches are well suppressed thanks to the double-MZ configuration as discussed in Fig. 4. Although the directional-coupler intersection was designed to have a 100% coupling ratio at a specific operating wavelength, the ratio is not 100% when the wavelength is different from the operating wavelength. Hence, (i) leakage first occurred at the intersection in each double-MZ switch. Next, (ii) the reflected light reached a successive directional-coupler intersection, and a second reflection occurred. Then, (iii) the twice-reflected light reached another intersection in another double-MZ switch, and the reflection from the other intersection was added. In this manner, the reflections accumulated and became the leakages observed at the output ports. We believe that crosstalk improvement could be achieved by replacing the directional-coupler intersections with broadband intersections .
3.4 Polarization-diversity configuration and 32-Gbaud DP-QPSK WDM signal transmission
We prepared two 8 × 8 switches that had almost the same transmission characteristics and created a polarization-diversity configuration by using fiber-based polarization beam splitters (PBSs) [14, 15]. The polarization-dependent loss (PDL) and differential group delay (DGD) of the 8 × 8 polarization-diversity switch were measured using an optical component analyzer (N7788B, Agilent), and the results are presented in Fig. 8. The PDL was approximately 0.5 dB in the 1525–1550 nm wavelength range. Note that the element switches in front of the output ports were adjusted so as to balance the insertion loss between the two switches and to minimize the PDL. The PDL degradation at wavelengths longer than 1550 nm was attributed to the difference between the wavelength dependences of the MZ switches on each switch chip. The DGD was less than 8 ps in the C-band, which is attributable to the residual difference between the lengths of the polarization-maintaining fibers after polarization splitting. The DGD can be compensated for by digital signal processing in digital-coherent transmission .
Figure 9 illustrates the experimental setup used for 32-Gbaud DP-QPSK four-channel WDM signal transmission over the switch. The switch state was that shown in Fig. 6(a), i.e., the state in which the signal was exposed to the most severe crosstalk from the other seven paths. Four tunable laser diodes with wavelengths of 1530.33 nm, 1535.04 nm, 1539.77 nm, and 1544.53 nm were connected to a 1 × 4 coupler. In the coupler, the lights from the laser diodes were mixed together. The mixed lights were modulated into a 32-Gbaud DP-QPSK signal by using an arbitrary waveform generator and a DP IQ modulator. The polarizations of the modulated lights were scrambled; then, the lights were led to the PBS connected to port 8 of both switch chips. The modulated lights were separated into two orthogonal polarizations, then launched into the switch through the polarization-maintaining high-delta fiber arrays edge-coupled to the chips, where the polarizations of both of the launched light were TM-like mode. The input optical power was 6.0 dBm/ch. To evaluate the signal qualities in a fully loaded condition of the switch in which all of the inputs are used, dummy cw lights at the same wavelength were launched into ports 1–7 in the same manner. The two output signals from port 7’ of the two switch chips were led to an output PBS, and the two polarizations were mixed. The output signals were amplified by an Er-doped fiber amplifier (EDFA), and one of the channels was selected by a tunable filter. The optical signal-to-noise ratio (OSNR) of the filtered light was adjusted by using an amplified spontaneous emission (ASE) light source and a variable optical attenuator. The OSNR-adjusted light was led to a coherent receiver consisting of a polarization- and phase-diversity optical receiver circuit, balanced photo detectors, and a 33-GHz real-time oscilloscope. The received electric signal was demodulated by an offline digital signal processor (DSP) employing a 2 × 2 butterfly adaptive equalizer.
Figure 10 presents the measured Q-factors of the four wavelength channels versus the OSNR at the input of the coherent receiver. The Q-factors of the signals after going through the 8 × 8 switch agree well with those measured in the back-to-back case. This agreement indicates that the signals were transmitted through the 32-Gbaud DP-QPSK WDM without any penalties. The penalty-less transmission is also confirmed by the constellations observed at an OSNR of 20 dB, which are depicted in Fig. 11. The symbols are clearly separated in all of the constellations and agree well with those in the constellation measured under the back-to-back condition at a wavelength of 1535.04 nm.
We demonstrated the broadband operation of a silicon photonics 8 × 8 PILOSS switch. Double-MZ-element switches were used to improve the crosstalk and expand the low-crosstalk bandwidth. The fabricated 8 × 8 switch exhibited a fiber-to-fiber insertion loss of 11.2 dB and a bandwidth wider than 30 nm for −20 dB crosstalk. A polarization-diversity 8 × 8 switch composed of two switches achieved a low PDL of less than 0.5 dB. Furthermore, we demonstrated 32-Gbaud DP-QPSK WDM signal transmission without significant degradation. We believe that these results indicate that double-MZ-switch-based PILOSS switches are promising for use as optical fiber path switches in future energy-efficient network systems. Achieving on-chip polarization diversity  is the next step toward developing a more compact and cost-effective optical switch.
Ministry of Education, Culture, Sports, Science, and Technology (MEXT), Japan; Project for Developing Innovation Systems.
This work was partly supported by the Project for Developing Innovation Systems of MEXT, Japan. The device fabrication was supported by the TIA Super Clean-Room of AIST. The authors are grateful to Furukawa Electric Co., Ltd. for providing the high-delta polarization-maintaining fiber.
References and links
1. J. Kurumida, K. Ishii, A. Takefusa, Y. Tanimura, S. Yanagimachi, H. Takeshita, A. Tajima, K. Fukuchi, H. Honma, W. Odashima, H. Onaka, K. Tanizawa, K. Suzuki, S. Suda, K. Ikeda, H. Kawashima, H. Uetsuka, H. Matsuura, H. Kuwatsuka, K. Sato, T. Kudoh, and S. Namiki, “First demonstration of ultra-low-energy hierarchical multi-granular optical path network dynamically controlled through NSI-CS for video related applications,” in European Conference on Optical Communication (Optical Society of America, 2014), paper PD.1.3. [CrossRef]
2. S. Nakamura, S. Takahashi, M. Sakauchi, T. Hino, M. Yu, and G. Lo, “Wavelength selective switching with one-chip silicon photonic circuit including 8 × 8 matrix switch,” in Optical Fiber Communication Conference (Optical Society of America, 2011), paper OTuM2. [CrossRef]
4. T. Tanemura, L. Langouche, and Y. Nakano, “Strictly non-blocking 8 × 8 silicon photonic switch based on optical phased array,” in European Conference and Exhibition on Optical Communication (IEEE, Valencia, 2015), paper P.2.9. [CrossRef]
5. T. J. Seok, N. Quack, S. Han, R. S. Muller, and M. C. Wu, “Large-scale broadband digital silicon photonic switches with vertical adiabatic couplers,” Optica 3(1), 64–70 (2016). [CrossRef]
6. K. Suzuki, K. Tanizawa, T. Matsukawa, G. Cong, S.-H. Kim, S. Suda, M. Ohno, T. Chiba, H. Tadokoro, M. Yanagihara, Y. Igarashi, M. Masahara, S. Namiki, and H. Kawashima, “Ultra-compact 8 × 8 strictly-non-blocking Si-wire PILOSS switch,” Opt. Express 22(4), 3887–3894 (2014). [CrossRef] [PubMed]
7. K. Tanizawa, K. Suzuki, M. Toyama, M. Ohtsuka, N. Yokoyama, K. Matsumaro, M. Seki, K. Koshino, T. Sugaya, S. Suda, G. Cong, T. Kimura, K. Ikeda, S. Namiki, and H. Kawashima, “Ultra-compact 32 × 32 strictly-non-blocking Si-wire optical switch with fan-out LGA interposer,” Opt. Express 23(13), 17599–17606 (2015). [CrossRef] [PubMed]
8. K. Tanizawa, K. Suzuki, S. Suda, H. Matsuura, K. Ikeda, S. Namiki, and H. Kawashima, “Silicon photonic 32 × 32 strictly-non-blocking blade switch and its full path characterization,” in 21st Optoelectronics and Communications Conference / International Conference on Photonics in Switching Proceedings (online) (IEEE Photonics Society, Niigata, 2016), paper PD2–3.
9. T. Goh, A. Himeno, M. Okuno, H. Takahashi, and K. Hattori, “High-extinction ratio and low-loss silica-based 8 × 8 strictly nonblocking thermooptic matrix switch,” J. Lightwave Technol. 17(7), 1192–1199 (1999). [CrossRef]
10. K. Suzuki, K. Tanizawa, S. Suda, H. Matsuura, K. Ikeda, S. Namiki, and H. Kawashima, “Broadband 8 × 8 Si-wire PILOSS switch with double Mach-Zehnder switch elements,” in European Conference and Exhibition on Optical Communication (VDE, Düsseldorf, 2016), paper W.4.P1.SC2.13.
11. Y. Shoji, K. Kintaka, S. Suda, H. Kawashima, T. Hasama, and H. Ishikawa, “Low-crosstalk 2 x 2 thermo-optic switch with silicon wire waveguides,” Opt. Express 18(9), 9071–9075 (2010). [CrossRef] [PubMed]
12. S. Suda, K. Tanizawa, K. Suzuki, H. Matsuura, K. Ikeda, S. Namiki, and H. Kawashima, “Fast and accurate calibration method for large-port-count Si-wire PILOSS optical switch,” in Asia Communications and Photonics Conference 2015 (Optical Society of America, Hong Kong, 2015), paper AS4A.2. [CrossRef]
13. Y. Ma, Y. Zhang, S. Yang, A. Novack, R. Ding, A. E. J. Lim, G. Q. Lo, T. Baehr-Jones, and M. Hochberg, “Ultralow loss single layer submicron silicon waveguide crossing for SOI optical interconnect,” Opt. Express 21(24), 29374–29382 (2013). [CrossRef] [PubMed]
14. K. Tanizawa, K. Suzuki, S. Suda, K. Ishii, J. Kurumida, G. Cong, T. Inoue, K. Ikeda, S. Namiki, and H. Kawashima, “Off-chip polarization-diversity 4 × 4 Si-wire optical switch with digital DGD compensation,” IEEE Photonics Technol. Lett. 28(4), 457–460 (2016). [CrossRef]
15. K. Suzuki, K. Tanizawa, S. Suda, H. Matsuura, T. Inoue, K. Ikeda, S. Namiki, and H. Kawashima, “Silicon-photonics polarization-insensitive broadband strictly-non-blocking 8 × 8 blade switch,” in Optical Fiber Communication Conference and Exhibition (OFC) (Optical Society of America, Los Angeles, 2017), paper Tu3F.5.
16. K. Suzuki, K. Tanizawa, S. H. Kim, S. Suda, G. Cong, K. Ikeda, S. Namiki, and H. Kawashima, “Polarization-rotator-free polarization-diversity 4 × 4 Si-wire optical switch,” IEEE Photonics J. 8(2), 1 (2016). [CrossRef]