OECC/PS2016

Stacked Wavelength Selective Switch Design for Low-cost CDC ROADMs

Haining Yang1, Brian Robertson1, Peter Wilkinson1,2, and Daping Chu1,2 * 1 Roadmap Systems Ltd, St John’s Innovation Centre, Cowley Road, Cambridge CB4 0WS, UK

2 Department of Engineering, University of Cambridge, 9 JJ Thomson Avenue, Cambridge CB3 0FA, UK

* daping.chu@roadmapsystems.co.uk, dpc31@cam.ac.uk

Abstract: A highly flexible stacked switch module is proposed, where 48 independent flex-spectrum 1×12 WSSs can be realised on a single 4k LCOS device. This design can handle all the switching operations within a CDC ROADM. Keywords: Wavelength Selective Switch (WSS), Reconfigurable optical add/drop multiplexer (ROADM)

I. INTRODUCTION

A wavelength selective switch (WSS) is the key building block for a modern reconfigurable optical add/dropmultiplexer (ROADM) [1], which enables reconfigurable optical networks. A typical WSS is able to selectively route individual wavelength division multiplexed (WDM) channels entering its input fibre port to any of the output fibre ports according to the software configuration that is remotely controlled by the service providers. In recent years, phase-only liquid crystal on silicon (LCOS) spatial light modulators (SLMs) have become the technology of choice for WSSs, due to their software upgradable nature and support for flexible spectrum switching [2]. Efforts [3-6] have also been made to improve the port count, crosstalk levels and passband shapes in LCOS WSSs. WSSs are usually based on the ‘disperse-and-select’ optical design, where the WDM channels from the input port are diffracted along the dispersion axis at the LCOS plane, before being switched to the target output ports according to the sub-holograms displayed on the corresponding areas of the LCOS device. Due to the limited number of pixels available on the current generation LCOS devices, anamorphic optics are invariably used in these designs to convert the input signals to elongated beams at the LCOS plane. Correspondingly, output ports are arranged along a switching axis that is orthogonal to the dispersion axis. Although such an approach is able to increase the port count in one axis, it fails to fully exploit the two dimensional (2D) nature of the LCOS pixel array. Moreover, for such a configuration, all the undesirable diffraction orders due to the LCOS quantization effects will also appear along this switching axis, which makes it fundamentally difficult to suppress crosstalk, especially in WSSs with high port counts.

In this paper, we proposed a stacked WSS module, which does not use anamorphic optics. It utilises 2D beam steering and can incorporate 48 independent 1×12 flex-spectrum WSSs on a single 4k LCOS device within one module. This module can also be configured to realise 12×12 contentionless wavelength cross-connect (WXC), for a low-cost, small-footprint add/drop solution for colourless, directionless and contentionless (CDC) ROADMs.

II. OPTICAL DESIGN

The principle of the proposed stacked module is shown in Fig. 1, which depicts M 1×N WSSs (M=3 and N=8 in thisexample) plus an array of objective lenses (LA), a relay system (L1 and L2) and DEMUX optics (Pg). Each of the 1×N WSSs has a fibre array cluster, which consists of 1 input, N output fibre ports and the corresponding micro-lens array. These clusters are arranged along the y-axis, each acting as an independent 1×N WSS, with each input light beam illuminating a spatially distinct row of sub-holograms (e.g., S1, S2, and S3). The WDM input is launched into each WSS via the central fibre in the corresponding cluster. The objective lens generates a beam waist of radius ωo at plane Po. The

Fig. 1. Design principle for the stacked WSSs based on a single LCOS device, (a) side view, (b) top view and (c) system view.

ME2-3

Copyright@IEICE

OECC/PS2016

4f relay system images this beam waist at the SLM plane (PSLM). The static grating (Pg) imparts an angular displacement of βG(λ) to each wavelength channel in the x-z plane. Signal beams associated with these wavelengths illuminate separate sub-holograms displayed on the LCOS device. The sub-hologram width can be adjusted to enable flex-spectrum switching. The sub-hologram for a wavelength channel could be a grating of period T, orientated at an angle of φ with respect to the local xy-coordinate system, that diffracts the light beam such that it leaves the LCOS SLM with a propagation vector of k(ρ, φ, λ), where ρ is the angle of the vector with respect to the local z-axis. The diffracted beam is subsequently imaged at Po by the relay system. The LA in the output optics shown in Fig. 2 converts the propagation vector of a wavelength channel, k(ρ, φ, λ), to a beam position that is offset from the optical axis. The angle is controlled such that the beam is concentric with respect to the intended output fibre, thereby maximising coupling efficiency as shown in Fig. 2(a). A secondary lenslet array, LF, focuses the wavelength channels into the output fibre array.

Our development work has shown that it is possible to use 50×50 pixels for each 50GHz channel within the C-band. By using a standard 4k LCOS device (such as the Jasper JD2704 with 4096x2400 pixels), 48 independent WSSs can be stacked along the y-axis of a single chip. In the absence of anamorphic optics, the un-modulated input signal to each sub-hologram has a circular beam shape on PSLM, which is designed to cover 31×31 pixels, and hence achieve the 4th order super Gaussian passband shape. The small number of pixels covered by the beam limits the maximum grating period to 10 pixels, which is larger than the minimum period of 7 pixels required to realise sufficient switching efficiency and reasonably low crosstalk level for two switchable positions along a given axis. The circular beam on PSLM allows 2D steering, giving 8 switchable output ports arranged on a Cartesian grid, as shown in Fig. 2(b). The number of the switchable output ports can be further increased, given the same beam steering range, to 12 if the fibre ports are arranged in a hexagonal pattern, as shown in Fig. 2(c).

Fig. 2. (a) Output optics design; 2D beam steering over the fibre ports arranged on (b) a Cartesian grid and (c) a hexagonal grid.

III. CROSSTALK SUPPRESSION

Crosstalk from a blazed grating on an LCOS occurs due to phase level quantization and pixel fringing fields whereby the electric field due to the voltage applied to a pixel leaks towards neighbouring pixels. The optical replay field incorporating the edge effect is shown for a blazed grating with a period of 7 pixels in Fig. 3(b), where a series of discrete diffraction orders can be seen. These may coincide with a number of fibre ports in the conventional LCOS WSS design, with a one-dimensional linear fibre port array and anamorphic optics, leading to high crosstalk levels. In the design illustrated in Fig. 1, however, only the -1st diffraction order of the blazed grating can coincide with an un-targeted port, in either the Cartesian or hexagonal arrangement. In this specific case, the -1st diffraction order will cause a crosstalk of about -20 dB, which can be suppressed by using advanced computer generated hologram algorithms for a specific order [5]. More effectively, we propose a general hardware approach, with the same principle as the wavefront encoding technique [4], to suppress crosstalk by building an optical asymmetry into the system via a matched array of axicon spatial filters positioned at plane Po. Specifically, each of the lenslets LA shown in Fig. 2(a), has an associated axicon element of wedge angle β, as shown in Fig. 3(a). In operation a Gaussian beam enters the switch via the central fibre of each fibre cluster, and is relayed to plane Po, where the axicon phase element imparts a radially linear phase delay. The wavefront leaving the axicon is imaged by lenses L1 and L2 at the LCOS plane, with each wavelength channel being separated by the DEMUX grating such that it illuminates a separate sub-hologram. If no phase pattern were displayed on a sub-hologram, the light would simply reflect off the LCOS device, and be re-imaged at the axicon element. Due to the phase reversal of light on reflection, the overall action of the axicon is to double the radially linear phase shift, causing the light to be focused by LA into a ring pattern centred about the input fibre. As a result, for a specific propagation vector, k(p,φ,λ) the sub-hologram must also display a phase pattern that compensates for the axicon filter in a lock-and-key approach that maximizes coupling efficiency for the +1st order. Consequently, all other mth diffraction orders are focused into ring patterns with a radius dependent on (m-1), where m is the diffraction order. Fig. 3(c) illustrates the corresponding sub-hologram and an optimized axicon filter (wedge angle of β = 0.38°), which reduces the -1st order crosstalk power by 11.8 dB, with only a penalty of 0.25 dB on insertion loss.

Fig. 3. (a) Illustration of an axicon operation; (b) replay field using a blazed grating with a period of 7 pixels with no filter; (c) equivalent replay field

using an axicon matched spatial filter at plane Po with a wedge angle of β = 0.38°.

Copyright@IEICE

OECC/PS2016

IV. APPLICATIONS

The large number of independent WSSs in this module allows various potential applications. In particular, this module offers two possible low-cost and small-footprint CDC add/drop solutions for ROADMs. Fig. 4(a) shows the current CDC add/drop solution based on multicasting switches including splitters and space switches. Although such a multicasting switch has the advantages of small footprint and low cost, it is not scalable due to the insertion loss in the splitters. Consequently, amplifier arrays are required to maintain a sufficient signal level, which increases the cost and power consumption. The second solution shown in Fig. 4(b) replaces the splitters with an array of WSSs so that the insertion loss no longer scales with the add/drop port count. However, the large number of WSS modules required here significantly increases the cost and footprint. This conundrum can be solved by using our stacked WSSs module in an M 1×N WSS configuration, as shown in Fig. 4(c). Our stacked WSSs module is able to efficiently route WDM channels in this configuration, with minimum increase in the cost and footprint due to sharing of common optics and the LCOS device between its 48 independent WSSs. As shown in Fig. 4(d), our stacked WSSs module can be further used as a WXC, which is able to realise contentionless wavelength switching between multiple input and output ports. In this case, space switches are no longer required to achieve CDC add/drop, further increasing the system integration. The operational principle of the WXC is given in Fig. 4(e) based on an example of a 4×4 WXC, in which the output ports are paired between WSSs. In this specific example, each WSS still has 8 spare ports for adding or dropping wavelengths to or from corresponding directions. Our module is able to realise 12×12 WXC using 24 of the independent WSSs. The rest 24 1×12 WSSs within our module can be either used to construct a second 12×12 WXC or deployed as standard 1×12 WSSs at the transit part of ROADM, according to the needs of the network operator. In the latter case, all the switching operations within a ROADM could be handled by our stacked WSS module.

Fig. 4. CDC add/drop solutions for ROADM: (a) multicasting switch; (b) WSS array; (c) stacked WSSs module in M 1×N configuration; (d) stacked

WSSs module in M×N configuration; (e) 4×4 WXC based on 8 WSSs.

V. CONCLUSIONS

We have proposed a stacked WSS module, in which 48 independent 1×12 flex-spectrum WSSs can be realised on a single 4k LCOS device. The proposed module can maximise the switching capability in terms of port count per pixel, and minimise the cost and energy consumption per port. The crosstalk in such WSSs is primarily due to -1st diffraction order of the sub-holograms, which can be reduced by building an optical asymmetry into the optical system, via an array of axicon matched spatial filters. Our results show an 11.8 dB reduction in the crosstalk with only a 0.25 dB insertion loss penalty. This proposed module is highly flexible and it can be configured as an array of 1×N WSSs, an N×N WXC with N up to 12 or any combination thereof. This offers a low-cost and small-footprint switching and CDC add/drop solution for ROADMs.

REFERENCES

[1] T. A. Strasser, and J. L. Wagener, “Wavelength-selective switches for ROADM applications,” IEEE J. Sel. Top. Quantum Electron. 16(5), 1150-1157 (2010).

[2] S. Frisken, G. Baxter, D. Abakoumov, H. Zhou, I. Clarke, and S. Poole, “Flexible and grid-less wavelength selective switch using LCOS technology,” in Optical Fiber Communication Conference/National Fiber Optic Engineers Conference, paper OTuM3 (2011).

[3] K. Suzuki, Y. Ikuma, E. Hashimoto, K. Yamaguchi, M. Itoh, and T. Takahashi, “Ultra-High Port Count Wavelength Selective Switch Employing Waveguide-Based I/O Frontend,” in Optical Fiber Communication Conference, paper Tu3A.7 (2015).

[4] B. Robertson, Z. Zhang, H. Yang, M. M. Redmond, N. Collings, J. Liu, R. Lin, A. M. Jeziorska-Chapman, J. R. Moore, W. A. Crossland, and D. P. Chu, “Reduction of crosstalk in a colourless multicasting LCOS-based wavelength selective switch by the application of wavefront encoding,” Proc. SPIE 8284, 82840S (2012).

[5] H. Yang, B. Robertson, and D. Chu, “Crosstalk reduction in holographic wavelength selective switches based on phase-only LCOS devices,” Optical Fiber Communication Conference, paper Th2A.23 (2014).

[6] C. Pulikkaseril, L. A. Stewart, M. A. F. Roelens, G. W. Baxter, S. Poole, and S. Frisken, “Spectral modeling of channel band shapes in wavelength selective switches,” Opt. Express 19(9), 8458-8469 (2011).

Copyright@IEICE