Spatial and Planar Optical Circuit (SPOC) Technology and Its Application to Photonic Network Devices
Because photonic networks have evolved from point-to-point systems to ring or mesh networks, higher scalability is required in the devices used in optical nodes. Hybridization of waveguide and free-space optics or spatial and planar optical circuits (SPOCs) may provide the necessary solutions to meet this requirement. A SPOC platform is attractive because it can take advantage of both waveguide technology and free-space optics. Waveguide technology provides a high degree of integration of optical functionality for such devices as splitters and non-wavelength selective switches, while free-space optics supplies a high degree of parallelism with two-dimensional spatial light modulators such as liquid crystal on silicon (LCOS) devices. In this article, we summarize the basics of SPOC technology and review its application to reconfigurable optical add/drop multiplexing (ROADM) devices. The key element of a waveguide on a SPOC platform is the spatial beam transformer, which has the same circuit structure as an arrayed waveguide grating but functions as a microlens array and provides attractive features such as dense integration of switches. An LCOS device has numerous phase modulating pixels, enabling flexible manipulation of lightwaves. We used a SPOC platform to construct and demonstrate several types of wavelength selective switches for ROADM applications.
Keywords: spatial and planar optical circuit (SPOC), wavelength selective switch, free-space optics
As photonic networks have evolved from point-to-point systems to ring or mesh networks, reconfigurable optical add/drop multiplexing (ROADM) systems have been deployed, which are required to have colorless, directionless, and contentionless (CDC) capabilities . Optical switching devices such as wavelength selective switches (WSSs) and multicast switches (MCSs) are important components in CDC-ROADM systems. Although waveguide-based WSSs were first proposed in early ROADM networks [2, 3], free-space optics based configurations have mainly been deployed in the actual systems only since the end of the last decade. The free-space optics based approach enables us to provide not only a sharp wavelength filter due to a high spatial resolution but also port scalability due to a high degree of parallelism. The MCS, on the other hand, is constructed based on waveguide technology  since it does not provide any wavelength-selective operation, which usually requires a bulk grating component located in the free space.
Combining waveguide technology with free-space optics can provide a solution that is not achievable using either method individually. The use of an arrayed waveguide grating (AWG) in combination with a MEMS (microelectromechanical systems) mirror was first proposed in 2001 , and since then, several ideas including the use of a wavelength blocker [6, 7], tunable optical dispersion filter [8–13], and WSS [14–17] have been reported.
In this article, we describe our study, which is focused primarily on WSSs, and explain our concept of a spatial and planar optical circuit (SPOC) that combines the waveguide and free-space optics to get the best of both technologies.
2. SPOC 
An example of the SPOC concept is illustrated in Fig. 1. The waveguide frontend is used as an input/output (I/O) component for interfacing with the fiber optics and free-space optics, and traditional bulk optics including a grating, lenses, and spatial light modulator (SLM) are used in the free space to complete the design. The diffraction grating needed to implement the WSS is located with its dispersion axis aligned normal to the waveguide.
One unique and essential circuit element in the waveguide is a spatial beam transformer (SBT) , as shown in Fig. 2(a). The SBT consists of multiple I/O waveguides, a slab waveguide, and arrayed waveguides. The SBT has a similar layout to that of a conventional AWG, but the path length difference of the arrayed waveguides is set to zero. It functions as follows. A lightwave input from the center input waveguide (indicated by the dashed arrow in Fig. 2(a)) spreads in the slab waveguide, couples to the arrayed waveguides, and exits from the chip facet to free space. Arrayed waveguides with a uniform path length difference maintain the wavefront of the lightwave so that a plane wave is output from the facet. Changing the input to another waveguide (indicated by a solid or dotted arrow) tilts the wavefront, so the output wavefront is also tilted.
Furthermore, it is also possible to bend the wavefront by controlling the path length among the arrayed waveguides so that one can launch not only a collimated beam but also a converging (or diverging) beam. The design flexibility in the SBT enables us to remove the discrete microlenses in the front of the fiber array. The equivalent optical design using conventional optics is shown in Fig. 2(b) for comparison.
The integration of (micro)lenses means that the SBT is suitable for packaging the ports densely. Moreover, the beams are expected to be positioned well because there is no misalignment with the microlenses. The uniformity of the launched beam angle over 96 SBT circuits fabricated using the standard silica-based planar lightwave circuit (PLC) process is shown in Fig. 3. Excellent pointing accuracy less than +/-0.1 degrees was obtained for both polarizations, which is three times smaller than in the typical fiber-array-based frontend. The launch angle is translated into the transmission wavelength of a conventional AWG. Adjustment of the transmission wavelength in an AWG is well established, so the launch angle in the SBT can also be controlled well.
Some applications using SPOC are described in the following section.
3. Applications of SPOC
A typical node configuration for a CDC-ROADM network is shown in Fig. 4. The building blocks of each node consist of a WSS, a transponder aggregator (TPA), and optical amplifiers. Some of the wavelength division multiplexing (WDM) signals arriving at a node are routed by the WSS in different directions. The remaining signals are dropped and added at the node. They are directed from/to the TPA to/from the WSS.
3.1 Ultra-high port count WSS 
Port scalability is important for WSSs used in CDC-ROADM nodes because each node has to deal with a large number of WDM channels. For example, for an 8-degree node with 88 wavelength channels, the total number of WDM channels is 704. If 8 × 8 MCSs are used as the TPA, a WSS with 95 output ports (88 for the TPA; 7 for the other directions) is needed to achieve a drop rate of 100%.
The 1 × 95 WSS optical design using the SPOC technology is shown in Fig. 5(a). This ultra-high port count WSS consists of a waveguide frontend, a collimating lens, polarization diversity optics (polarization beam splitter (PBS) and half-wave plate (HWP)), a diffraction grating, port-selecting and focusing cylindrical lenses, and an LCOS-based SLM.
The WDM signal input to the waveguide frontend is radiated into free space. The signal is collimated with a collimating lens in a direction normal to the waveguide substrate (y-axis). Because the LCOS is a polarization-sensitive device , we implemented polarization diversity optics in which the signal is split into two orthogonal linear polarizations along the y-axis. One of them is rotated 90 degrees with the HWP so that the linearly polarized signal is incident to the LCOS. The signal then passes through a diffraction grating of which the dispersion direction is along the y-axis. Next, the signal is collimated in the x-axis direction by the port-selecting lens and focused along the y-axis by the focusing lens. The SLM reflects the signal back along the same route to the frontend in the y-axis while it steers the signal so that it hits a different position on the frontend in the x-axis. Hereafter, we call the x- and y-axes the switching and dispersion axis, respectively.
The SBT circuit was designed to have a beam radius of 26 µm at the chip facet with a port-to-port separation of 117 µm; therefore, the total effective aperture and physical chip size in the x-axis are only 11.2 and 17.5 mm, respectively. With the help of the high-dispersion grating, the respective focal lengths of the port-selecting and focusing lenses are 150 and 100 mm. The transmission spectra of the 1 × 95 WSS, in which each of the 95 WDM channels was routed to different output ports, is shown in Fig. 5(b).
3.2 Low-loss TPA 
Another key device for constructing a CDC-ROADM node is a TPA; to date, a PLC-based MCS has been the only commercial solution. A block diagram of the MCS is shown in Fig. 6(a). Incoming WDM signals are first broadcast by N, which is the number of transponders connected to the MCS. M units of the 1 × N splitters are integrated in one chip, where M is the number of degrees in the ROADM node. N units of M × 1 switches are also integrated to select one route of M degrees to send to a transponder.
Although the silica-based PLC has been matured and has proved to have high reliability, the MCS suffers from a principle loss of splitting. Therefore, it is becoming impractical with the scaling up in the number of transponders.
Another approach to implement the M × N TPA is shown in Fig. 6(b). Here, M units of 1 × N WSSs and N units of M × 1 switches are connected. Unlike the MCS, each output port is connected to only one input port at a time; this is a necessary and sufficient functionality for a TPA because each transponder handles a single wavelength channel at a time. Since it has no splitters inside and therefore no intrinsic loss, this WSS-based TPA is preferable for a large scale TPA with large N compared to a conventional MCS-based TPA.
Using the SPOC platform enables the frontend of the M units of 1 × N WSSs and N units of 1 × M switches to be integrated in one chip, as shown in Fig. 7(a). The M × 1 switch can be made with Mach-Zehnder interferometers in a tree arrangement with TO (thermo-optics) phase shifters, as shown in Fig. 7(b). The diagram in Fig. 7(a) illustrates the case for M = 4 and N = 4. WDM signals from four different directions are input from the ports labeled M fiber direction and multiplexed in the angular domain by SBTs. They hit different areas on the LCOS, and each one supports a different WSS. This means that multiple WSSs can be angularly multiplexed.
The measured spectra of the fabricated 8 × 24 TPA are shown in Fig. 8, where 24 WDM signals of the 50-GHz-grid channel spacing in the C-band were independently routed to 24 output ports. The 24 graphs correspond to the output ports, and each plot has eight curves representing the input ports.
We can see directionless and contentionless switching from the set of graphs for output ports 1, 5, 9, and 13, where the same wavelengths from different input ports were routed to different output ports. Wavelength selective switching is clearly evident in the set of graphs for output ports 2, 6, 10, and 14, where four different channels from input port 3 were routed to different output ports. Flexible-grid operation is also clearly evident in the set of graphs for output ports 4, 8, 12, and 16, showing respectively that 50-, 100-, 150-, and 200-GHz transmission bands were obtained. The minimum insertion loss was 10.7 dB, which is lower than the theoretical loss of 13.8 dB of an MCS with the same port count.
3.3 Integrated wavelength cross-connect with add/drop functionality
If we look again at Fig. 4, we can see many WSSs in the wavelength cross-connect (WXC) area, which is indicated by the dashed lines. WSS multiplexing is also useful for constructing a WXC.
Schematic diagrams of the functional block and optics system for the integrated WXC we constructed are shown in Fig. 9 . We used a broadcast-and-select type WXC configuration in which 1 × N splitters broadcast WDM signals from the ingress ports, and each N × 1 WSS selects one of them. The drop and add ports are used to access the TPAs. Because it is illogical to return a signal in the direction from which it originated, the splitter and the WSS for the same direction are not connected. The output ports of the splitters and the input ports of the WSSs that are not connected are used as drop and add ports. As shown in Fig. 9(b), N units of 1 × N splitters can be integrated into the waveguide frontend, and one of the split ports of each splitter is looped back to serve as a drop port. An SBT is integrated at the bottom of the waveguide frontend for use in adding ports.
The functionality of the constructed WXC was experimentally tested for channel bandwidths of 100, 150, and 200 GHz to evaluate the flexible-grid operation. Routing to Egress 5 as an example connection between the Ingress and Egress ports is shown in Fig. 10(a). The plot shows that every input port was routed to every output port except for the ports with the same port number in each bandwidth. The connections for the Add and Egress ports are shown in Fig. 10(b). Each Add port was routed to the corresponding Egress port without a splitter, so the insertion losses of the Add ports were lower than those of the Ingress ports. The losses from the Ingress to the Egress ports included an intrinsic splitting loss of 9 dB, which is inevitable due to the broadcast-and-select configuration.
We have described platform technology that combines two-dimensional planar waveguide photonics and three-dimensional free-space optics, which we call a spatial and planar optical circuit, or SPOC. The planar waveguide technology provides a high degree of integration of optical functionality for such devices as microlenses, splitters, and simple switches, while free-space optics supplies a high degree of parallelism with two-dimensional spatial light modulators such as LCOS devices.
With the unique advantages of SPOC, we demonstrated three applications for ROADM devices in a photonic network. These include the ultra-high port count 1 × 95 WSS, the low-loss 8 × 24 TPA, and the single module 8 × 8 WXC.