A novel liquid crystal on silicon (LCOS)-based wavelength selective switch (WSS) is proposed, fabricated, and demonstrated. It employs a multilayered arrayed waveguide grating (AWG) as a wavelength multiplex/demultiplexer. The LCOS deflects spectrally decomposed beams channel by channel and switches them to desired waveguide layers of the multilayered AWG. In order to obtain the multilayered AWG with high yield, phase errors of the AWG is externally compensated for by an additional phase modulation with the LCOS. This additional phase modulation is applied to the equivalent image of the facet of the AWG, which is projected by a relay lens. In our previously-reported WSS configuration, somewhat large footprint and increased cost were the drawbacks, since two LCOSs were required: one LCOS was driven for the inter-port switching operation, and the other was for the phase-error compensation. In the newly proposed configuration, on the other hand, both switching and compensation operations are performed using a single LCOS. This reduction of the component count is realized by introducing the folded configuration with a reflector. The volume of the WSS optics is 80 × 100 × 60 mm3, which is approximately 40% smaller than the previous configuration. The polarization-dependent loss and inter-channel crosstalk are less than 1.5 dB and −21.0 dB, respectively. An error-free transmission of 40-Gbit/s NRZ-OOK signal through the WSS is successfully demonstrated.
©2013 Optical Society of America
A reconfigurable optical add-drop multiplexer (ROADM) is a key subsystem for the node in the next-generation wavelength division multiplexing (WDM) network since it can efficiently process a huge amount of information, replacing the conventional energy-consuming electrical routers. In such a ROADM, a considerable number of wavelength selective switches (WSSs) are required for connecting the optical paths between add/drop modules and I/O fibers toward the other nodes . Therefore, a WSS should be compact and low-cost (e. g. low optical complexity and high yield) from the viewpoint of installation space and cost for the node. Furthermore, a flexible-grid operation function  is strongly desired for the WSS, in which center frequencies and spacings between the channels can be dynamically changed with discrete frequency steps. This function enables us to allocate the spectral resources efficiently for WDM channels which have various bandwidths and formats.
A WSS mainly consists of a wavelength multiplex/demultiplexer (MUX/DEMUX), Fourier optics, and a switching engine, as shown in Fig. 1(a). Typically, a liquid crystal on silicon (LCOS) device [3, 4] or a micro-electromechanical systems (MEMS) mirror array [5, 6] is employed for the switching engine, and a bulk grating [3–5] is used for the MUX/DEMUX. A WSS operates as follows: 1) the WDM signal input to the WSS is spectrally decomposed by the DEMUX, Fourier-transferred by the Fourier optics, and focused onto the switching engine; 2) the switching engine, which consists of a number of switch units segmented along the x-axis, deflects the decomposed wavelength components channel by channel and directs them toward any of the output ports; and 3) the deflected wavelength components are spectrally recombined by the MUX. By changing the number of neighboring switch units directing the lights toward the same output port, the spectral grid for each channel is flexibly allocated. Compared to the MEMS mirror array, the LCOS is an advantageous engine for such a flexible-grid operation. This is because the minimum frequency step in allocating the grid for a channel is proportional to the width of the switch unit w (namely, the pixel size in the LCOS, or the width of a single mirror in the MEMS), and the LCOS has one order finer spatial resolution (w~10 μm) than that of the MEMS array (w~100 μm) .
In the conventional LCOS-based WSS with a bulk grating [3, 4], a complicated beam-shaping optics and a micro-lens array were typically required to form a MUX/DEMUX, as shown in Fig. 1(a). This is because the beam profiles along the x- and y-directions need to be asymmetrically designed to obtain the satisfactory performances of the WSS: The y-directional profile is designed to reduce the diffraction loss and the truncation loss at the LCOS, while the x-directional profile is designed to obtain flat-top spectra. As for the y-directional design, the wavefront of each spectrally decomposed light is modulated to a blazed shape at the LCOS for the beam-deflection. The maximum deflection angle of the light θmax is limited due to the diffraction loss, which is induced by the pixelization effect of the wavefront . Typically, the inter-port distance along the y-direction Dport is large (~1000 μm) since it is determined by the spatial pitch of the I/O fiber array. Therefore, a focal length of the Fourier optics F should be sufficiently long so that one can switch the light to the distant output ports with small θmax. A micro-lens array is introduced to adjust the spot size of the beam after the input port in the y-direction and mitigate the excess divergence of the beam during the long free-space propagation over F. The y-directional spot size on the LCOS is therefore, designed to be comparable to the height of the LCOS in order to reduce the truncation loss. Regarding the x-directional design, the spot size of the focused monochromatic light at the LCOS determines the spectral shape of a passband for a channel [5, 7]. The x-directional spot size of the monochromatic light should be narrow on the LCOS in order to obtain the ideal flat-top spectra and maintain the quality of the signal passing through the WSS. A beam-shaping optics comprising cylindrical or anamorphic elements was typically used to control the spot size in x-direction. For the above reasons, the conventional LCOS-based WSS requires many bulky components (e. g. the micro-lens array, beam-shaping optics, and bulk grating) and resulted in a high assembly cost. In particular, an ultra-precise alignment is needed for the micro-lens array, and it consumes excessive time for calibration.
We have proposed LCOS-based WSSs employing a multilayered AWG for MUX/DEMUX [9–14] in order to simplify the assembly and reduce the cost. A schematic of our early WSS configuration [9, 11] is shown in Fig. 1(b). In this configuration, several AWGs fabricated on separate substrates are stacked and bonded together to form the multilayered AWG. Each waveguide layer corresponds to the I/O port of the WSS. Compared to the conventional WSS with bulk-grating, this AWG-based WSS has a simpler MUX/DEMUX configuration with less bulky components. This is because the required beam-profile in the x-direction can easily be obtained by arranging the layout of the arrayed waveguides. However, we still need a cylindrical micro-lens array for controlling the beam size in the y-direction. This is due to the distant layout of the I/O ports (Dport~1000 μm), which is limited by the thick substrate of the waveguide. Moreover, phase error of the AWG leads to a large insertion loss and crosstalk of the WSS. Since the phase errors are fluctuated among the waveguide layers, the yield of the multilayered AWG significantly decreases when the layer count is large for a high-port-count WSS. The conventional phase-error trimming scheme with UV irradiation  is not applicable for the multilayered waveguides.
In our subsequently-reported WSS configuration, we have introduced the following two schemes to remove the micro-lens array and to compensate for the phase error of the AWG. First, we have developed a monolithically integrated multilayered AWG , where each waveguide layer is closely (Dport~25 μm) fabricated on a common substrate. This close arrangement of the I/O ports in the y-direction enables the inter-port signal switching with small θmax and short F without using the micro-lens array, as shown in Fig. 1(c). Second, in order to obtain the multilayered AWG with high yield, we introduced an additional LCOS into the WSS [12, 14]. This additional LCOS modulates the wavefront of the relayed image of the facet of the AWG, and it can externally compensate for the distorted wavefront induced by the phase errors of the AWG layer by layer. Although this compensation scheme greatly boosts the yield of the multilayered AWG, the drawbacks of the WSS are somewhat large footprint and increased cost since we need two LCOS devices.
In this paper, we propose an improved WSS design, where the folded configuration with a reflector is newly introduced. With this configuration, both switching and phase error-compensation operations are performed with a single LCOS. In the subsequent sections, structures of the proposed WSS and a principle of the switching and phase-error compensation are described. Based on the parameter design aiming at the practical WSS, we have assembled the WSS by using the monolithic multilayered AWG. A phase-error compensation and an error-free transmission of 40-Gbit/s non-return-to-zero on-off keying (NRZ-OOK) signal are demonstrated.
2. Structure and principle
2.1 Conventional non-folded structure
Figures 2(a) and 2(b) show the schematics of AWG-and-LCOS-based WSS configurations: (a) previously-reported structure employing two LCOSs [12, 14]; and (b) newly proposed WSS driven by a single LCOS with a folded configuration. The previously-reported WSS consists of the multilayered AWG, a pair of LCOSs (LCOS-1, LCOS-2), a Wollaston prism polarization beam splitter (PBS), a half-waveplate (HWP), and three lenses (Lens-1, Lens-2, and Lens-3), as shown in Fig. 2(a). The Fourier-transforms (FTs) are executed three times by the lenses until the image on the facet of the AWG is transferred onto the plane-C. Since this imaging operation is equivalent to a single Fourier-transform of the AWG’s facet, a spectrally decomposed light appears on the plane-C. LCOS-1 is placed on the plane-C and is driven for deflecting the beams in the y-direction for inter-port switching. The additional LCOS (LCOS-2) is placed on the plane-B, where the relayed image of the facet of the AWG is projected by two lenses (Lens-1 and Lens-2). This relayed image consists of several beams corresponding to the waveguide layers in the multilayered AWG. The wavefronts of these relayed beams are distorted in the x-direction due to the phase error of the AWG, as shown in Fig. 3. This wavefront distortion causes the excess insertion loss and crosstalk by deforming the beam-profiles on the LCOS-1 in the x-direction. The phase errors in the AWG deviate among the waveguide layers. Therefore, the phase errors need to be compensated for layer by layer in order to obtain the WSS with high performance operation. By modulating the wavefront of each relayed beam with individual phase pattern at LCOS-2, we can compensate for the phase error for all the layers .
A polarization-diversity scheme with the PBS and HWP is employed to compensate for the large polarization-dependent loss (PDL) of the LCOSs. In this scheme, the PBS splits the image on the plane-A into two images corresponding to the polarization modes. One of the split images passes the HWP to rotate its polarization direction by 90°. Therefore, all the images on the LCOSs have the same polarization states. The relayed beams of the AWG’s facet on the plane-B are spatially split corresponding to the polarization mode of the AWG. Therefore, we can individually compensate for the phase errors for both polarization modes by LCOS-2. This technique is useful when the waveguide exhibits birefringence and the phase errors are polarization-dependent .
2.2 Proposed folded structure
We can fold the WSS optics with a reflector utilizing the symmetries of the optics: folding once at plane-B and twice at the center planes of the Lens-2 and Lens-3, as shown in Figs. 2(b) and 4(a). In this folded configuration, the plane-B and -C are coplanar, and the single LCOS is located on this plane. The upper part of the LCOS covers the plane-B and is driven for the phase-error compensation, and the lower part is assigned to the plane-C and is driven for the switching operation, as shown in Fig. 4(b). During the imaging operations from the plane-A to -C, the x-polarization mode of the AWG passes the HWP twice, while the y-polarization mode passes the HWP once, as shown in Fig. 4(a). Therefore, the polarization state of all the images on the LCOS is aligned to the x-polarization, and hence, this WSS is polarization-insensitive. The longitudinal axes of the Lens-1 and Lens-2' are off-centered by H, so that the planes-A, -B, and -C are spatially isolated in the y-direction.
The optical behaviors in the y-z plane during the switching operation are detailed in Fig. 4(a). The incident light launched from the input port, that is, one waveguide layer of the multilayered AWG, is Fourier-transformed by Lens-1 and collimated onto the plane-A, where the PBS is located. The light is split by the PBS according to its orthogonal polarization state. Then, the split light at the plane-A is Fourier-transformed again by Lens-2', and focused onto the plane-B on the LCOS. The spacing between the focal positions of the polarization-split beams on the plane-B is determined by designing the beam-splitting angle θPBS of the PBS. The reflected lights from the plane-B are Fourier-transformed thirdly by Lens-2', and are collimated onto the plane-C. Two beams which are split by the PBS cross each other on this plane. The crossing angle of these two collimated lights on the plane-C is equal to θPBS. The blazed-shaped phase modulation by the LCOS is then applied to the light on the plane-C, as shown in the inset of the Fig. 4(a). The first-order diffraction light is deflected with an angle of θ (|θ|≤θmax). Then, the deflected lights go back to Lens-2' and focused on the offset position to the focus of input light on the plane-B. On the plane-B, the optical modes of the output layers are projected by the relay-optics (Lens-1 and Lens-2'). The deflected beams couples to any of these relayed image of the output modes. The input x-polarization mode (Ex, in) couples to the y-polarization mode of the output waveguide (Ey, out), and the input y-polarization mode (Ey, in) couples to the x-polarization mode of the output waveguide (Ex, out), as shown in Fig. 4(b). Then, the reflected lights from the plane-B of the LCOS are polarization-multiplexed and relayed to the facet of the AWG passing through Lens-2', HWP, PBS, and Lens-1. Finally, the light reaches to the target output waveguide layers in the multilayered AWG.
2.3 Phase error identification and compensation
In order to compensate for the phase error of the multilayered AWG, it is required to identify the value of the phase error. We have developed an in situ calibration scheme with the WSS  to identify and compensate for the phase error of the multilayered AWG. In this calibration scheme, the phase modulation patterns on the plane-B of the LCOS are expressed by the combination of the Legendre orthogonal polynomials. The optimal set of Legendre coefficients that reduces the insertion loss is calculated with an iterative trial-and-error-based optimization algorithm. This method is convenient since it does not require any external instruments such as a wavefront sensor. The calibration flow of this scheme is detailed in Section 5.1.
3. Design of the WSS
3.1 Target performances
We designed the WSS aiming at the practical usage, as detailed in Table 1. The x-directional design is shown in Fig. 3, and y-directional design is shown in Fig. 4(a). The configurations in the x- and the y-direction were independently designed. The x-directional design determines the spectral performances, such as the operation spectral range, spectral flatness, and spectral granularity in adjusting the channel grid for flexible-grid operation. The y-directional design determines the port count, PDL, and the diffraction loss at the LCOS. We tried to reduce the size of the WSS as possible, satisfying the target specifications shown in Table 1.
3.2 y-directional design
First, we describe the y-directional design referring to Figs. 4(a) and 4(b). Assuming that all beams in the WSS optics have Gaussian profiles, the y-directional spot size, i. e. the 1/e2-diameter of the intensity profile, of the beam at plane-C ωCy is expressed as
The port count of the WSS Nport is determined byFig. 4(a). 2F1yθmax indicates the maximum moving distance of the beam on the facet of the AWG. The function ‘floor’ is the round-down function. According to Eqs. (1)–(3), Nport is expressed by8]. When the pixel size h of the LCOS in the y-direction is 8.0 μm, and the allowable diffraction loss is 0.06 dB, θmax is calculated to be ~0.7°. In this calculation, we neglected the diffraction loss induced by pixel-gaps and the field-fringing effect . Therefore, further diffraction loss is anticipated in the actual device.
Figure 5 shows the calculated port count of the LCOS as a function of Dport and ω0y, when A = 0.96, θmax = 0.7 °, and HC = 5.7 mm. In the conventional LCOS-based WSS employing the bulk grating [3, 4] or multilayered AWG with stacked substrates [9, 11] (Dport~1000 μm), the micro-lens array is necessary to expand the spot size after the MUX/DEMUX (ω0y~140 μm) and to obtain the sufficient port count (Nport~10). On the other hand, the lens array-free (ω0y~4 μm) 1 × 10 WSS is feasible if we fabricate the multilayered waveguide with Dport of less than 29.0 μm. Therefore, monolithic fabrication  is necessary to form the waveguide layers closely with each other. A directional coupling between the waveguide layers leads to an inter-port crosstalk. This undesired crosstalk is negligibly small if we design Dport to be larger than 3ω0y.
Figures 6(a)–6(c) show the ray-tracing calculation result for the WSS, where ω0y = 4.0 μm, Dport = 26.4 μm, Nport = 10, and HC = 5.7 mm. With the ray-tracing method, we can approximately calculate the optical paths and beam profiles in the WSS optics. Optical paths with all the switching state and both polarization modes are overlaid. All 22 beams from the AWG (1 input × 10 outputs, x- and y- polarization modes) are correctly projected on the plane-B and -C, as shown in Figs. 6(a) and 6(b). The arrangement of the PBS and HWP (H, θPBS, hHWP, and L, shown in Fig. 4(a)) is optimized so that the WSS provides the minimum PDL. The worst PDL is exhibited at the optical paths which are coupling to the top waveguide layer and coupling to the bottom waveguide layer (path-I, -II, -III, and -IV). In order to reduce the PDL, the optical path-I should pass the HWP, and the other paths (path-II, -III, and -IV) should not pass the HWP. The PDL is calculated by performing the overlap integral between the input optical mode and output mode on the surface of the HWP, and the beam profile of each mode is obtained with the ray-tracing. The calculated worst PDL with this design is less than 0.34 dB, which satisfies the requirement shown in Table 1. The height of the LCOS is designed to be 14.4 mm (8 μm × 1800 pixels), which is a sufficient height to cover all the beams on plane-B and -C without excess truncation loss.
3.3 x-directional design
Next, we design the x-directional configuration in order to obtain the target spectral response of the WSS, as listed in Table 1. Since the operation spectral range of the WSS should be smaller than the free-spectral range (FSR) of the AWG, the diffraction order m of the AWG is determined asFigure 7(a) shows the phase modulation pattern on plane-C when certain bandwidth σ is allocated for a WDM channel. The focal position of the monochromatic light on the plane-C xf(λ) is dependent on the wavelength of the light. xf(λ) is expressed asEq. (5), m is limited to be less than 44, when the WSS is operated in C-band (λc = 1547.5 nm, ΔλRange = 35 nm). The higher m leads to a larger diffraction loss of the AWG at the edge channels in the C-band. However, the too small m increases F1x and the size of the WSS, according to Eq. (6). Therefore, we determined the optimal m as 27.
Wch(σ) is the total width of the switch units which are deflecting the light to direct it to the monitored output port, as shown in Fig. 7(a). The flatness of the spectra is determined by the ratio between Wch(σ) and the spot size of the monochromatic light ωCx [5, 7]. Wch(σ) is equal to the moving distance of the light when the input frequency is shifted by σ. According to Eq. (6), Wch(σ) is expressed asFig. 7(a)) [5, 7]. The term A(x, λ) is expressed byEquation (10) indicates that the spectral shape is determined only by the variable Wch(σ)/ωCx when σ is fixed. We evaluate the spectral shape using the 3-dB pass bandwidth of the spectrum. The wider 3-dB bandwidth represents the ideal flat-top spectra. Figure 7(b) shows a calculated 3-dB pass bandwidth as a function of Wch(σ)/ωCx, when σ is set to 100 GHz. In order to obtain a sufficient flatness of the spectra with a 3-dB pass bandwidth of 90.5 GHz, we determine the optimal Wch(σ)/ωCx as 2.87.Eqs. (5) and (6). Therefore, Narray is determined to be 358.
The spectral granularity Δν in adjusting σ for the flexible-grid operation is determined by the pixel size w of the LCOS in the x-direction. When Wch is increased by w (1 pixel), σ is increased by Δν. According to Eq. (7), w and Δν are given by
The width of the active area on the LCOS should be larger than the spot size of the beams in plane-B in order to reduce the truncation loss. Also LCOS should cover all the wavelength components in the C-band, which are decomposed on the plane-C. The width of the LCOS of 7.2 mm (8 μm × 900 pixels) satisfies these requirements.
Table 2 summarizes the designed parameters for the WSS components discussed in this section.
4. Monolithically integrated multilayered AWG
4.1 Fabrication of the AWG
According to the simulated results in Section 3, each waveguide layer needs to be fabricated close to each other (Dport≤29.0 μm) in order to obtain a WSS with port count of 10. We have developed a monolithic fabrication technique , with which several AWGs are closely fabricated on a single substrate. Figure 8(a) describes the procedure for fabricating the multilayered waveguide: (i) A waveguide-core of the bottom layer on a silica substrate is deposited, and the waveguide pattern is fabricated with reactive ion etching (RIE); (ii) over-cladding is deposited with chemical vapor deposition (CVD), and the waveguide is annealed; (iii) Surface of the over-cladding layer is flattened with chemical mechanical polishing (CMP); (iv) The second-layer waveguide core is deposited and patterned; (v) The over-cladding for the second-layer waveguide is deposited, and the waveguide is annealed; (vi) By iterating the above processes (iii)–(v), we obtain the AWG with higher layer count.
As a preliminary trial, we have fabricated a monolithic multilayered AWG with two waveguide layers. Figure 8(b) shows a microscopic image of the facet of the fabricated AWG. The measured inter-layer distance Dport and beam-diameter in the y-direction ω0y were ~26.4 μm and ~3.9 μm, respectively. The relative refractive index difference of the waveguide was 2.5%. The size of the fabricated AWG chip was 27 × 13.5 mm2.
4.2 Evaluation of fabricated AWG
We measured the spectral response of the fabricated multilayered AWG. By analyzing the shapes of the spectra, we can roughly estimate the magnitude of the phase error in the AWG. We connected the fiber array to the I/O ports of the AWG and attached a flat mirror to the facet of the arrayed waveguide. Sweeping the wavelength and polarization state of the input light, we measured the reflected optical power using the circulator.
Figures 9(a) and 9(b) show the measured loss spectra of the fabricated two-layered AWG: (a) the AWG in the bottom layer named as Layer-1; and (b) AWG in the top layer named as Layer-2. The power difference between the black solid and the red broken curves represents the PDL, and each spectrum exhibited a large PDL. This large PDL can be explained by a combination of two spectra: one is a spectrum of the x-polarization mode, and the other is that of the y-polarization mode, as shown in the insets in Figs. 9(a) and 9(b). The peak-wavelength of each spectrum was different between the polarization modes and between the waveguide layers. Further, each spectrum was significantly deformed with undesired side-lobes. These results of spectral shift and deformation indicate that the fabricated AWG had large phase errors, and the value of the phase error was peculiar to each polarization mode and waveguide layer. These large phase errors led to the enlarged insertion loss and crosstalk by deforming the beam profiles at the plane-C corresponding to each polarization mode in each waveguide layer. Therefore the phase errors need to be compensated for.
5.1 Assembly and phase error compensation
We assembled the 1 × 1 WSS using the fabricated two-layered AWG. Figure 10 shows the photograph of the assembled WSS. The volume of the optics was 80 mm (width) × 100 mm (length) × 60 mm (height). Size reduction of approximately 40% was achieved, compared to the previous configuration with two LCOSs .
We compensated for the phase-errors of the AWG with the assembled WSS. Figure 11 shows the setup for the phase-error compensation. The input port of the Layer-1 was connected to the tunable laser through the polarization controller. The output port of the Layer-2 was connected to the power meter. The tunable laser emitted a CW light, and its wavelength was fixed at 1547.713 nm. The wavefront of the light on the plane-C was locally modulated by the LCOS with a blazed shape, so that the deflected light coupled to the output waveguide layer. The width of this modulated pattern Wch on the plane-C was set to 2 pixels, which was nearly equal to the target spot size of the beam ωCx on the plane-C. Four beams were projected onto the plane-B, which corresponded to the two waveguide layers and two polarization modes of the AWG. The wavefronts of these four beams were modulated with individual phase patterns with the LCOS ϕlayer, pol(x). The first subscript of ϕ(x), ‘layer,’ indicates the layer number of the AWG (1 or 2); and the second subscript, ‘pol,’ indicates the x- or y-polarization. The optical transmission of the input x-polarization light was calibrated with ϕ1, x(x) and ϕ2, y(x), and the transmission of the input y-polarization light was calibrated with ϕ1, y(x) and ϕ2, x(x). Each phase pattern on the plane-B was expressed by a combination of Legendre polynomials, and optimal set of the Legendre coefficients that reduced the transmission loss was calculated with a trial-and-error based optimization algorithm . Thus, ϕlayer, pol(x) was expressed as17]. Second, we fixed the incident polarization state to y-polarization, and found the optimum alayer, pol for ϕ1, y(x) and ϕ2, x(x) with the PSO-based method. During the optimization, each profile of the monochromatic light on the plane-C corresponding to each propagation mode of the AWG was concentrated inside the blazed shaped phase pattern.
5.2 Spectral response
Figures 12 show the measured spectral response of the WSS, where the phase patterns for interleaving operations with 200-GHz channel spacing were encoded on the LCOS. Figure 12(a) shows the results without the phase-error compensation (wavefront on the plane-B was not modulated), and Fig. 12(b) shows the results with the phase error compensation (wavefront on the plane-B was modulated). The blue curves show the transmission when odd channels are switched and even channels are blocked; and the green curves show the transmission vice versa when the even channels are switched and odd channels are blocked. The power difference between the solid and broken curves represents the PDL. The inter-channel crosstalk was defined as the extinction ratio at the center of each channel. One can clearly see that the spectral characteristics were greatly improved by compensating for the phase error of the AWG. The insertion loss was reduced from ~40.0 dB to ~29.0 dB, the worst PDL was reduced from 3.2 dB to 1.5 dB, and the worst crosstalk was reduced from −5.8 dB to −21.0 dB. These results indicate that the deformed profiles of the monochromatic light on the plane-C were correctly reshaped by the phase modulations on the plane-B for each waveguide layer and each polarization mode. We also demonstrated the flexible-grid operations with the WSS. Figure 13 shows the measured spectra of the WSS when σ was changed from 100 GHz to 200 GHz. By increasing Wch with a step of 2 pixels, the transmission bandwidth was successfully increased by 25 GHz without generating any spectral ripples. When σ was set to 100 GHz, the measured 3-dB pass bandwidth was 80.8 GHz, which was somewhat narrower than the designed value of 90.5 GHz. This spectral narrowing and the residual PDL of less than 1.5 dB were mainly due to a misalignment of the optics. The insertion loss of ~29.0 dB was attributed to the loss of the AWG (8.1 dB), the loss of the LCOS (8.3 dB), loss of the free-space optics (4.7 dB), and other loss induced by the misalignment (7.9 dB), as shown in Table 3. These spectral narrowing, PDL, and insertion loss will be further suppressed by introducing well-aligned low-aberration optics, the AWG with improved design, and an LCOS with high reflectance.
5.3 Signal transmission
We transmitted a 40-Gbit/s non-return-to-zero on-off keying (NRZ-OOK) signal through the WSS and evaluated a quality of the output signal. Figure 14 shows an experimental setup for the signal transmission. First, we generated the signal by driving a transmitter, which was composed of a wavelength-tunable laser (TL) and Mach-Zehnder modulator (MOD). The signal format was 40-Gbit/s NRZ-OOK with a pseudo random bit sequence of 231−1. The carrier frequency of the signal was set to 193.7 THz (λ = 1547.715 nm). Then, the signal light is launched into the WSS. The allocated bandwidth for the channel σ was set to be 100 GHz, and the center frequency of the passband was set to be equal to the carrier’s one. The WSS switched the signal to the output port with the selected bandwidth σ. The output signal after the WSS was guided to the receiver, which consisted of a variable optical attenuator (VOA), an Erbium-doped fiber amplifier (EDFA), a band pass filter (BPF), a photo detector (PD), and a bit error rate tester (BERT). We measured the bit error rate (BER) characteristics, adjusting the optical signal-to-noise ratio (OSNR) with the VOA and EDFA. The BPF was used for removing the redundant amplified spontaneous emission (ASE) noise of the received signal. The BPF had a flat-topped spectral shape with a 3-dB pass bandwidth of 938.6 GHz. This pass bandwidth was wide enough to correctly evaluate the filtering characteristics of the WSS. Each instrument was connected through short fibers, whose total length was about 30 m. Therefore, the chromatic dispersion in the fibers was negligibly small.
Figure 15(a) shows measured BER characteristics with 40-Gbit/s NRZ-OOK signal, where the results under two conditions are compared: (i) back-to-back transmission; and (ii) through the WSS. One can see that both BER curves exactly fitted each other without OSNR penalties. The eye diagrams at the OSNR of 23.4 dB are shown in Fig. 15(b). The eyes were clearly open for both conditions. These results indicate that the filter response of the WSS was flat-top enough to transmit the 40-Gbit/s signal without degrading the signal quality.
A novel compact AWG-and-LCOS-based WSS was proposed, fabricated, and demonstrated. The multilayered AWG was used for the MUX/DEMUX, and the LCOS was employed for the inter-port switching. The phase error of the multilayered AWG was externally compensated for by applying the additional phase modulation onto the LCOS, and the yield of the WSS was greatly improved. Our previously-reported WSS required two LCOS devices and resulted in high cost and large size. On the other hand, by employing a folded configuration with a reflector into the WSS, both switching and phase error compensation operations were performed on the single LCOS. The size of the optics was 80 mm × 100 mm × 60 mm, which was 40% smaller than that of the previously-reported WSS. Such a low-cost, compact, and high yield WSS is useful for constructing the next-generation ROADM, where many WSSs are required. By compensating for the phase error of the AWG for each waveguide layer and each polarization mode, the insertion loss was dramatically improved from 40.0 dB to 29.0 dB. The WSS also exhibited good spectral responses, where the inter-channel crosstalk was less than −21.0 dB and the PDL was less than 1.5 dB. Also, the optical signal modulated with 40-Gbit/s NRZ-OOK was successfully transmitted through the WSS without OSNR penalties.
This work was supported in part by “Project for Developing Innovation Systems of the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan,” and by “International Communications Research Grant from KDDI foundation.”
References and links
1. Y. Li, L. Gao, G. Shen, and L. Peng, “Impact of ROADM Colorless, Directionless, and Contentionless (CDC) Features on Optical Network Performance,” J. Opt. Commun. Netw. 4(11), B58–B67 (2012). [CrossRef]
2. S. Poole, S. Frisken, M. Roelens, and C. Cameron, “Bandwidth-flexible ROADMs as Network Elements” Proc. OFC/NFOEC 2011, OTuE1 (Los Angeles, USA, 2011).
3. M. A. F. Roelens, S. Frisken, J. A. Bolger, D. Abakoumov, G. Baxter, S. Poole, and B. J. Eggleton, “Dispersion Trimming in a Reconfigurable Wavelength Selective Switch,” J. Lightwave Technol. 26(1), 73–78 (2008). [CrossRef]
4. Y. Sakurai, M. Kawasugi, Y. Hotta, M. D. S. Khan, H. Oguri, K. Takeuchi, S. Michihata, and N. Uehara, “LCOS-Based Wavelength Blocker Array with Challel-by-Channel Variable Center Wavelength and Bandwidth,” IEEE Photon. Technol. Lett. 23(14), 989–991 (2011). [CrossRef]
5. D. M. Marom, D. T. Neilson, D. S. Greywall, C.-S. Pai, N. R. Basavanhally, V. A. Aksyuk, D. O. López, F. Pardo, M. E. Simon, Y. Low, P. Kolodner, and C. A. Bolle, “Wavelength-selective 1 × K Switches Using Free-Space Optics and MEMS Micromirrors: Theory, Design, and Implementation,” J. Lightwave Technol. 23(4), 1620–1630 (2005). [CrossRef]
6. D. M. Marom, C. R. Doerr, M. Cappuzzo, E. Chen, A. Wong-Foy, and L. Gomez, “Hybrid Free-space and Planer Lightwave Circuit Wavelength-selective 1x3 Switch with Integrated Drop-side Demultiplexer” Proc. ECOC 2009, PD1.9 (Vienna, Austria, 2009).
7. P. Wall, P. Colbourne, C. Reimer, and S. McLaughlin, “WSS Switching Engine Technologies” Proc. OFC/NFOEC 2008, OWC1 (San Diego, USA, 2008).
8. D. Sinefeld and D. M. Marom, “Insertion Loss and Crosstalk Analysis of a Fiber Switch Based on a Pixelized Phase Modulator,” J. Lightwave Technol. 29(1), 69–77 (2011). [CrossRef]
9. K. Sorimoto, H. Tsuda, H. Ishikawa, T. Hasama, H. Kawashima, K. Kintaka, M. Mori, and H. Uetsuka, “A Compact High-Port-Count Wavelength Selective Switch Using LCOSs and a Multi-Stacked AWG” Proc. 21st Annual Meeting of the IEEE Lasers & Electro-Optics Society (LEOS)2008, TuCC2 (Newport Beach, USA, 2008).
10. K. Sorimoto, H. Tsuda, H. Ishikawa, T. Hasama, H. Kawashima, K. Kintaka, M. Mori, and H. Uetsuka, “Design of a Wavelength Selective Switch Using an LCOS and a Multi-stacked AWG Fabricated on Wedge-shaped Substrates” Proc. International Topical Meeting on Information Photonics (IP)2008, 3-4 (Awaji, Japan, 2008).
11. K. Sorimoto, H. Tsuda, H. Ishikawa, T. Hasama, H. Kawashima, K. Kintaka, M. Mori, and H. Uetsuka, “Demonstration of a Wavelength Selective Switch Using an LCOS and a Stacked Arrayed Waveguide Grating” Proc. ECOC 2009, P2.04 (Vienna, Austria, 2009).
12. K. Sorimoto, H. Tsuda, H. Ishikawa, T. Hasama, H. Kawashima, K. Kintaka, M. Mori, and H. Uetsuka, “Polarization Insensitive Wavelength Selective Switch Using LCOSs and Monolithically Integrated Multi-layered AWG” Proc. 15th OptoElectronics and Communications Conference (OECC), 6E2–4 (Sapporo, Japan, 2010).
13. K. Sorimoto, K. Kintaka, H. Kawashima, M. Mori, T. Hasama, H. Ishikawa, H. Tsuda, and H. Uetsuka, “1×6 Multicasting Operation in an LCOS-and-AWG-based Wavelength Selective Switch” Proc. 1st International Symposium on Access Spaces (IEEE-ISAS), GS3-B-3 (Yokohama, Japan, 2011).
14. K. Sorimoto, K. Kintaka, H. Kawashima, M. Mori, T. Hasama, H. Ishikawa, H. Tsuda, and H. Uetsuka, “Phase Error Compensation for Multilayered AWG in LCOS-based WSS,” IEICE Electron. Express 8(24), 2054–2060 (2011). [CrossRef]
15. K. Takada, T. Tanaka, M. Abe, T. Yanagisawa, M. Ishii, and K. Okamoto, “Beam-Adjustment-Free Crosstalk Reduction in 10 GHz-spaced Arrayed-Waveguide Grating via Photosensitivity under UV Laser Irradiation through Metal Mask,” Electron. Lett. 36(1), 60–61 (2000). [CrossRef]
16. U. Efron, B. Apter, and E. Bahat-Treidel, “Fringing-field Effect in Liquid-crystal Beam-steering Devices: an Approximate Analytical Model,” J. Opt. Soc. Am. A 21(10), 1996–2008 (2004). [CrossRef]
17. K. Sorimoto, K. Kintaka, H. Kawashima, M. Mori, T. Hasama, H. Ishikawa, H. Tsuda, and H. Uetsuka, “Fast Aberration-Correcting Algorithm for an SLM-based Optical Switch,” IEICE Electron. Express 7(23), 1728–1734 (2010). [CrossRef]