1. Introduction

Wavelength-division-multiplexing (WDM) technology has been widely deployed in broadband optical communication networks. To support greater network management, wavelength selective switching (WSS) based Reconfigurable Optical Add Drop Multiplexer (ROADM) technologies have been used to allow for cost-effective optically bypassing nodes. In the last decade, video based communications, cloud computing and social networking have become the dominant traffic sources in optical networks, and this has pushed optical carriers to increase not only the WDM channel rates but also the port counts, thus creating demand for cost-effective WSS-based ROADMs that enable better optical network flexibility and management. A WSS, as an emerging device with unprecedented performance and flexibility, can reduce the complexity of optical networks while enhancing the flexibility of the network nodes. It allows certain wavelength channels to be switched from any input port to any output port. An optical communication network based on WSSs has many attractive features, such as accelerating service deployment and capability of rerouting around a failure point, due to the ability of WSSs to automatically create optical paths for individual wavelength channels from one location to another.

WSSs have the ability to route a large number of wavelength channels efficiently in network nodes. However, most WSS modules of the present state-of-the-art use 1 × M WSSs, which switch wavelength channels between one input port and M output ports. Liquid crystal on Silicon (LCoS) WSSs were the first candidates and have been widely used to construct WSSs [1–3], based on its mature manufacture technology and excellent capability of optical beam steering. Shortly, the first MEMS-based WSS was demonstrated using a digital micromirror array as the spatial light modulator [4], and soon it became a popular solution [5–7]. More recently, some other approaches based on integrated optics technologies have also been proposed for the development of WSSs, including the use of the quantum-confined Stark effect [8], Raman amplifiers in conjunction with directional couplers [9] and thermooptic effects [10, 11]. Liquid-crystal-on-Silicon and MEMS-based WSSs are relatively bulky, however, they provide colorless, directionless and contentionless (CDC) operation, whereby the nodes using these WSSs can provide full flexibility in the allocation of resources, i.e., the flexibility to direct the ports of any wavelength to and from any fiber direction. An additional feature of these WSSs is their gridless operation [12, 13], i.e., their ability to adaptively allocate arbitrary wavelengths between the input and output ports, rather than using a fixed wavelength grid for channels.

Although the 1 × M WSS is beneficial for optical networks, as discussed above, it can only have a single input fiber port (or only one output port when it is used to route the WDM channels in the opposite direction). Furthermore, unlike N × M WSSs, the 1 × M WSS cannot offer non-blocking switching, i.e., it is incapable of routing different channels of similar wavelengths from different input ports to different output ports. On the other hand, N × M WSSs overcome the above-mentioned limitations of 1 × N WSSs and, hence, are highly desired in next-generation optical networks [14].

Most current N × M WSSs are constructed by connecting M × 1 and 1 × N WSS modules in series or parallel [12]. However, this discrete implementation of individual WSS modules becomes increasingly expensive when large numbers of 1 × N and M × 1 WSS modules are required. Very recently, hardware-compressed N × M WSS structures have been proposed, based on using MEMS, Liquid crystal on Silicon and planar lightwave circuit (PLC) technologies [13–15]. These WSS structures are extremely promising for next-generation high-connectivity, dynamic wavelength-routing optical networks.

In this paper, we propose and experimentally demonstrate a novel hardware-compressed, colorless, directionless, contentionless and gridless N × M WSS architecture capable of independently routing any wavelength channel from any input fiber port to any output fiber port. This structure also allows non-blocking wavelength selective switching between the N input ports and the M output ports.

2. N × M WSS architecture

The proposed N × M WSS architecture is based on the use of a 2D Opto-VLSI processor, which is a device capable of precisely steering optical beams incident on its surface along arbitrary directions. An Opto-VLSI processor comprises a silicon substrate, evaporated aluminum, a quarter-wave plate (QWP), liquid crystal (LC), Indium-Tin Oxide (ITO) and a glass. ITO is used as the transparent electrode, and evaporated aluminum as the reflective electrode. A voltage is applied between the ITO layer and the reflective electrode by the VLSI circuit below the LC layer to generate digital phase holograms for optical beam steering. Optically, the backplane is a flat dielectric mirror made of thin dielectric layers to eliminate the amplitude and optical path variations associated with the underlying aluminum pixel structure, and to minimize drop in electric field across the LC layer. Each pixel is able to be programmed to 256 discrete driving voltage states with a maximum input voltage of around 3.3 V, resulting in 256 discrete phase levels with a maximum optical phase change of 2π. The typical response time is around 10 ms. An Opto-VLSI processor is electronically controlled, software configurable, polarization independent, and very reliable since beam steering is achieved with no mechanically moving part [16].

Figure 1 illustrates the principle of the proposed N × M WSS. A fiber collimator array of N + M port count is used (N input ports and M output ports). The input signal from an input port is collimated and focused via a lens (Lens 1 with focal length f₁) onto a grating plate, which demultiplexes all the input signals along different directions for later selection and steering by a 2-D Opto-VLSI processor. Another lens (Lens 2 with focal length f₂), placed at a distance f₂ from the grating plate, is used to collimate the diverged optical beams in two dimensions, and map them onto the active surface of the Opto-VLSI processor (see Fig. 1(a)), which is partitioned into N + M rectangular pixel blocks. Each pixel block on the Opto-VLSI processor is associated to a specific input or output port. The pixel blocks associated to input ports are called input pixel blocks and the others are called output pixel blocks. A mirror, which is placed after the Opto-VLSI processor at a distance f₂ from Lens 2, is used to reflect the input signals (steered by the corresponding input pixel blocks) to the desired output pixel blocks, which further steer the signals into the corresponding output ports (steered by the associated output pixel blocks). As shown in Fig. 1, to switch a wavelength channel to an arbitrary output fiber port, two stages of beamsterring are needed, namely (1) steering the signal from the input port to the desired output pixel block, via the corresponding input pixel block and (2) after reflection off the mirror, steering the reflected beam through the output pixel block to the corresponding output fiber port. Both beamsteering processes are performed by uploading the appropriate phase holograms (or gratings) onto the relevant pixel blocks of the Opto-VLSI processor. By appropriately partitioning the active area of the Opto-VLSI processor into N + M pixel blocks, and sub-partitioning each pixel block into k sub-pixel-blocks (where k is the number of wavelength channels) any individual wavelength from any input port can independently be selected for switching and routing to any desired output port, thus realizing an N × M WSS.

 

Fig. 1 Proposed N × M WSS architecture.

Download Full Size | PDF

The proposed N × M WSS architecture, shown in Fig. 1, is simple and offers excellent flexibility to independently switching any selected wavelengths among different ports using computer generated holograms.

3. Experimental results and discussion

To demonstrate the principle of the proposed Opto-VLSI-based N × M WSS, a 2 × 3 WSS architecture was experimentally setup, as shown in Fig. 2. The fiber collimator array that collimated each input optical beam at about 0.5 mm diameter had ten ports with 1 mm spacing. Two ports were used as input ports (Port 4 and Port 5) and three other ports were used as output ports (Port 6, Port 7 and Port 8). Lens 1 and Lens 2 had focal lengths of 75mm and 100mm, respectively, and were placed at 75mm and 100 mm, respectively, from the grating plate, as illustrated in Fig. 2. A 256-phase-level two-dimensional Opto-VLSI processor, having 512 × 512 pixels and 15 µm pixel spacing, was used to independently and simultaneously select and switch the wavelength channels from an input fiber port to a desired output fiber port. A half wave plate was used to optimize the diffraction efficiency of the Opto-VLSI processor. An optical spectrum analyzer with 10 pm resolution and an optical switch were used to monitor the signals at the three output ports. A blazed grating, having 1200 lines/mm and a blazed angle of 36° at 1200 nm, was used to demultiplex the input WDM signals, which were then mapped onto the active window of the Opto-VLSI processor by Lens 2. A Labview software was especially developed to generate the optimized digital holograms that steer the desired wavelength channels and switch them to the desired output ports.

 

Fig. 2 Experimental setup to demonstrate the principle of the proposed N × M WSS architecture.

Download Full Size | PDF

The active window of the Opto-VLSI processor was divided into five pixel blocks corresponding and aligned to the five input and output ports. The size of each pixel block was 90 × 512 pixels so that the whole active window of the Opto-VLSI processor can accommodate all five pixel blocks (two input ports and three output ports). Optimized digital phase holograms were applied to the five pixel blocks, so that desired wavelength channels from the input signals illuminating the Opto-VLSI processor could be selected and switched into desired output ports. Wavelength selection was achieved by sub-partitioning the pixel blocks into sub-pixel-blocks and applying beamsteering phase holograms onto the sub-pixel–block, whereon the selected wavelengths were mapped. Note that the channel bandwidth for each channel selected for switching was defined by the width of the associated sub-pixel-block, thus it was possible to dynamically and independently control the individual channel bandwidths by simply reconfiguring the widths of the corresponding sub-pixel-blocks. Two Amplified Spontaneous Emission (ASE) light sources that cover C-band with power spectral density −16.7dBm/nm were used as input signals at Port 4 and Port 5. 27nm spectral width of the ASE light was mapped on the window of the Opto-VLSI processor (the other spectrum was out of the window). Therefore, each pixel accommodated a 27nm/512≈0.052nm bandwidth, which defines the WSS operating spectral resolution.

Figure 3(a) shows the output wavebands for a wavelength selective switching scenario where some wavebands of an input ASE signal launched into Port 4 were switched to Port 6, Port 7 and Port 8. Figure 3(b) shows the layout of the active window of the Opto-VLSI processor, displaying five pixel blocks (two for input ports and three for output ports) and the various phase holograms uploaded onto the sub-pixel-blocks for this wavelength selective switching scenario. The broadband optical signal input at Port 4 was mapped onto the bottom pixel block of the Opto-VLSI processor. To demonstrate the ability of the WSS to select and switch wavelength channels of arbitrary bandwidth, the ASE spectrum was divided into seven sub-pixel-blocks, each of which corresponds to a specific spectral bandwidth. The numbers of pixels for the sub-pixel blocks were chosen to be 58, 80, 58, 90, 58, 120 and 48, totaling 512 pixels. By driving the sub-pixel-blocks associated to Port 4 with appropriate phase holograms, seven wavelength channels mapped onto the bottom pixel block (associated to port 4) were steered to the pixel blocks associated to Port 6, Port 7 and Port 8, through reflection off the mirror (first-stage beamsteering). Particularly, two channels were steered to Port 6, two channels were steered to Port 7 and three channels were steered to Port 8. Each steering phase hologram was chosen to be the same size as its associated sub-pixel-block, which actually determined the bandwidth of the selected and switched wavelength channels. The steering phase hologram was a saw-tooth grating whose period was optimized to steer the beams for maximum coupling to the appropriate output fiber port. After first-stage beam steering, the optical beams steered by the various sub-pixel-blocks associated to Port 4 hit their target output sub-pixel-blocks, where a second stage of beamsteering was necessary to couple them into their associated output ports. It is important to note that the positions and sizes of the sub-pixel-blocks (and the uploaded steering phase holograms) associated with Port 6, Port 7 and Port 8 should be same as these of Port 4 to steer the desired wavelength channels to the desired output ports, as illustrated in Fig. 3(b). Figure 3(c) shows the overall spectral transmission from Port 4 to Port 6, Port 7 and Port 8, which demonstrates that the transmission is flat over the whole spectral range covering the Opto-VLSI window. Note also that in the above switching scenario no phase hologram was uploaded onto the pixel block associated to Port 5, though the input ASE signal was launched into Port 5.

 

Fig. 3 (a) Measured wavelength selective switch output power for switching from input Port 4 to output Ports 6, 7 and 8; (b) schematics of the corresponding beamsteering phase holograms uploaded onto the sub-pixel-blocks; (c) overall spectral transmission from input Port 4 to the output Port 6, 7 and 8.

Download Full Size | PDF

Figure 4 shows the output wavebands for another wavelength selective switching scenario, where the wavebands of the input ASE signal launched into Port 5 were switched to Port 6, Port 7 and Port 8 using appropriate phase holograms, while no steering holograms were uploaded onto the pixel block associated to Port 4. The input ASE signal was mapped onto the pixel block associated to Port 5, as shown in Fig. 4(b). Similarly, Fig. 4(c) shows the flat overall spectral transmission from Port 5 to Port 6, Port 7 and Port 8. The same sub-pixel-block sizes and positions used in the previous switching scenario (Fig. 3) were also used in this scenario; however, the various wavebands were switched to different output ports. For each waveband, a steering phase hologram was uploaded onto the corresponding sub-pixel-block to steer the waveband to the desired output pixel block, where another steering phase hologram steered it again and coupled it into the corresponding output port, as described in the previous paragraph. It is obvious from Figs. 3 and 4 that the proposed WSS architecture can arbitrary switch any wavelength channel of various bandwidth from Port 4 or Port 5 to Ports 6, 7 and 8.

 

Fig. 4 (a) Measured wavelength selective switch output power for switching from input Port 4 to output Ports 6, 7 and 8; (b) schematics of the corresponding beamsteering phase holograms; (c) overall spectral transmission from input Port 5 to the output Port 6, 7 and 8.

Download Full Size | PDF

To demonstrate the capability of the proposed WSS to simultaneously steer wavelength channels from both input ports to all output ports, another switching scenario was attempted. Figure 5 shows that the measured output power at Ports 6-8, with two arbitrary wavelength channels from both input Ports 4 and 5 being simultaneously switched to arbitrary output fiber ports. Figure 5(b) shows the schematics of the corresponding beamsteering phase holograms driving the various sub-pixel-blocks of the Opto-VLSI processor. Comparing Figs. 3, 4 and 5, it is obvious that negligible crosstalk is seen when all pixel blocks are operational.

 

Fig. 5 (a) Measured wavelength selective switch output power for simultaneously switching wavelength channels from input Port 4 and Port 5 to output Ports 6, 7 and 8; (b) schematics of the corresponding beamsteering phase holograms.

Download Full Size | PDF

Note that, the proposed WSS operates in a free space. The steered optical beams of the selected channels are diffracted by the pixel blocks and expand during propagation, resulting in passband effects. Compared to the wavelength selection method that needs only single-stage beam steering [16], the passband effect for this proposed WSS becomes non-ignorable. This might be because the Opto-VLSI processor was not placed at the focal plane of Lens 1. In order to investigate the passband effect, the size of each pixel block was reduced to 16 pixels, corresponding to a 0.84nm spectral width. Figure 6(a) shows the transmission spectra for simultaneously switching wavelength channels from Port 4 and Port 5 to Port 7. The schematics of the corresponding beamsteering phase holograms are illustrated in Fig. 6(b). As seen in Fig. 6(a), in addition to experiencing an extra power loss of around 6 dB, the transmission spectrum for each channel becomes “round”. This passband effect could be reduced if the mirror is replaced by a cylindrical mirror that reduces the beam divergence between the first and the second beamsteering stages.

 

Fig. 6 (a) Measured wavelength selective switch output power for simultaneously switching wavelength channels from Port 4 and Port 5 to Port 7 when the spectral width is about 0.84 nm (16 pixels); (b) schematics of the corresponding beamsteering phase holograms.

Download Full Size | PDF

Although principally the Opto-VLSI processor can be placed at any position between the mirror and Lens 2, a large challenge for its steering ability will be faced when it is too close to the mirror because large beam steering angles with high steering accuracy are needed for optical beam switching. However, a close distance between the Opto-VLSI processor and the mirror reduces the passband effects. Therefore, an Opto-VLSI processor with large-angle beamsteering capability can reduce the passband effects of the WSS.

Note that it is also possible to steer and couple the wavelength channels launched into an input port back into that same input port. This would increase the input and output port counts to 5 × 5. Moreover, the width of the active window of the Opto-VLSI processor determine the maximum spectral band that can be switched from the input ports to the output ports, while the height the active window controls the input and output port counts. Additionally, the use of an Opto-VLSI processor with a large window area and small pixel size enables a larger port count and higher wavelength selection resolution.

Since the switching of a wavelength channel from an input port to an output port requires two stages of beamsteering, the insertion loss of the WSS is relatively high compared to a conventional 1 × M WSS. The maximum measured total insertion loss of the WSS was 17 dB, which includes (i) the collimator coupling loss ~2 dB; (ii) the grating plate loss ~3 dB; (iii) lens aberration loss ~2 dB and (iv) the diffraction loss and insertion loss of the Opto-VLSI processor ~10 dB. These insertion losses can be minimized by optimizing the optical alignment and using high-quality optical components and coatings. However, it is important to mention that the two-stage beamsteering for each switching scenario results in lower crosstalk in comparison with a 1xM WSS. The measured crosstalk of the proposed WSS structure was less than −40 dB.

It is also important to note that in addition to the obvious colorless and contentless features, the proposed WSS structure enables directionless, non-blocking and gridless operations. The WSS is directionless because when Port 6, Port 7 and Port 8 were used as input ports, wavelength channel selection and switching to Port 4 and Port 5 was possible. Also, the WSS is non-blocking because wavelength channels of similar wavelengths can be independently steered to different output ports. Finally, the WSS enables gridless operation since the sizes of the sub-pixel-blocks can arbitrarily be adjusted to accommodate variable spectral bandwidths.

4. Conclusion

A novel Opto-VLSI-based N × M WSS structure has been proposed and experimentally demonstrated. Experimental results have shown that it is possible to select and switch any wavelength channel from any input optical fiber port to any output optical fiber port through a two-stage beamsteering process. A proof-of-concept 2 × 3 WSS structure has been developed, demonstrating colorless, directionless, contentionless and gridless operations and arbitrary wavelength switching with an insertion loss around 15 dB and crosstalk below −40 dB.