1. Introduction

Space-division multiplexing (SDM) either employing multiple fiber modes [1–3] or multi-core fibers [4] has recently seen significant attention as a means to use space as an additional dimension to overcome the fundamental bandwidth limits of single-mode fibers. However for SDM to become a viable solution for high bandwidth transmission, it is necessary to create SDM equivalents of the components and subsystems employed in today’s networks. In particular, wavelength selective switches (WSSs) [5] as the technology that underpins reconfigurable add-drop multiplexers (ROADMs) are a crucial part of modern wavelength routed networks. In order for networks to transition to SDM, ROADM solutions that are capable of switching spatial superchannels at different wavelengths to any of multiple outputs, need to be developed. A viable SDM-ROADM needs to be cost-efficient, be readily reconfigurable and compatible with modern techniques such as grid-less networking [6]. A WSS-based ROADM solution was previously demonstrated for multi-core-fiber (MCF) based networks [7]. Such a device consists of fanning-out the cores of the input and output MCFs and attaching them to a traditional single-moded WSS (SMF-WSS). All cores belonging to a single MCF are switched together as spatial superchannels and the performance of the device is consistent with the SMF-WSS on which it is based, as all spatial channels have the same Gaussian profile on the spatial light modulator (SLM) which serves as the switching element. A 9-port few-mode fiber WSS (FMF-WSS) which supported 3 spatial modes [8] based on an input/output array of few-mode fibers has also been demonstrated.

In the case of the MCF WSS, the fan-in/fan-outs provide a transition from the MCFs to standard SMFs creating a pseudo-single-moded device. For the FMF-WSS [8], no such transition occurs and the system is multi-moded along the entire optical path. As each mode occupies slightly different space on the SLM, each mode exhibits a different spectral passband [10,11]. Hence wavelength channels require larger guard-bands between them when compared to existing single-mode systems as the difference in the spectral response between the modes incurs mode dependent loss (MDL) and makes the outer portions of the band unusable leading to lower spectral efficiency. This effect becomes worse as more spatial modes are added and/or channels are more closely spaced in wavelength. Such a solution is not scalable to a large number of modes and does not support all the spectral filtering capability of traditional single-moded spatial light modulator (SLM) based WSSs. In this submission, we demonstrate a reconfigurable FMF-WSS based on single-mode fiber input/output arrays in conjunction with photonic lanterns [9]. The multi-mode to single-mode transition performed by the photonic lantern is analogous to the fan-in/fan-out adapters used in the MCF WSS [3] and enables the spectral performance of the switch to be consistent across all the spatial modes it supports (LP_0,1, LP_1,1a, LP_1,1b).

2. Principle

A traditional SMF-WSS [5] takes the form of Fig. 1(a). The input light from an SMF array is dispersed by a grating across the width of an SLM. Different phase modulations are then programmed onto the surface of the SLM such that different bands of wavelengths, correspond to different groups of pixel columns on the SLM, are tilted towards the desired output fiber. The simplest method of upgrading this scheme to support MDM is to replace the single-mode fiber array with a few-mode fiber array [8] as per the simplified diagram of Fig. 1(a). However as different regions of space on the surface of the SLM correspond to different wavelength bands, and as different modes do not occupy exactly the same space, the spectral passband of each mode is different [10,11]. An example of this is shown in Fig. 1(b) for the first 10 Laguerre-Gaussian modes (LP_0,1 to LP_3,1) incident on a common group of pixel columns on the SLM which is performing the switching and spectral filtering. Higher-order modes require more width on the SLM than the lower-order modes to achieve the same spectral extinction ratios and hence a conflict is created between densely packing the channels in wavelength and support for a high number of modes. Due to the more complicated spatial profiles of the higher-order modes, the shape of the passband itself also differs between the modes, which causes the MDL of a channel to roll-off before the insertion loss (IL), further limiting how closely the channels can be spaced in wavelength. In [8] this limited the spectral resolution to 50GHz.

 

Fig. 1 (a) Principle of operation behind FMF-based WSS [8] (b) Simulated example passbands for the first 10 LP modes. (c) Principle of operation of photonic lantern based WSS (d) Simulated example passbands for all modes are equal.

Download Full Size | PDF

To avoid this, the system presented here and summarized in Fig. 1(c), uses photonic lanterns [9] to convert all the light from the input/output few-mode fibers into multiple single-mode fibers attached to an SMF input/output array. After the lantern, all modes are spatially Gaussian and hence will all undergo the same spectral filtering by the SLM as per the example of Fig. 1(d). When a tilt is programmed onto the surface of the SLM it shifts the output beam relative to the input beam in the plane of the fiber array. This shift would typically correspond to the distance between the input and the desired output single-mode fiber for an SMF-WSS, because fibers in the input/output array at the same wavelength see the same shift. So for example, as shown in Fig. 1(c), coupling light at a particular wavelength from fiber 1 to 4 will also couple light from 2 to 5 and 3 to 6. Hence by attaching the single-mode fiber ports of photonic lanterns to adjacent cores of the input/output fiber array in a WSS, it is possible to switch between the different lanterns by programming tilts onto the SLM corresponding to shifts of multiples of N, where N is the number of spatial modes. This is the same basic principle employed for the MCF-WSS [7] except in that case, MCF fan-in/fan-out adapters took the place of the photonic lanterns and N is the number of cores rather than the number of spatial modes.

The device demonstrated here deviates slightly from the simplified diagram of Fig. 1(c). In this case, the input/output single-mode fiber array used actually consists of two columns of 24 fibers each as shown in Fig. 2(a). Two ports of the photonic lanterns are attached to one column of the array and the third ports are attached to fibers in the other column. Each column of fibers, labelled A and B in Fig. 2(a), is then addressed independently by the top and bottom of the same SLM to define the fiber array shifts Δx_A and Δx_B respectively. This is used to create a FMF-WSS which supports 11 FMF output ports where each port supports three spatial modes. In practice the SMF-WSS portion of the device is as per Fig. 2(b), with the 48 total SMFs for the two columns visible on the right of the image. 24 SMFs for column A will attach to two of the ports of the transmit and receive photonic lanterns and 12 SMFs from column B will attach to the remaining ports of the lanterns.

 

Fig. 2 (a) SLM switching element divided into two regions used to address each of the two columns in the fiber array. (b) Image of the single-moded WSS upon which the device is based.

Download Full Size | PDF

3. Experiment

The system is characterized using an SLM based reconfigurable spatial diversity optical vector network analyzer (SDM-OVNA) [12] which is summarized in Fig. 3(a). It is a swept-wavelength interferometer based technique which allows the complete mode transfer matrix of the system to be measured as a function of wavelength. This system operates on the same principle as a previously demonstrated SDM-OVNA [13]. However, instead of fixed beamsplitters and phase plates, it employs an SLM based mode-multiplexing system [14]. The fact that the system is implemented using an SLM rather than fixed optics, offers much greater flexibility and allows it to characterize the system in an arbitrary mode-basis. That is, it can be used to emulate butt-coupling of the system to an arbitrary fiber type. In this case, the SLM is used to generate the modes of a step-index 30μm core 0.06 numerical aperture (NA) few-mode fiber, which approximates the refractive index profile of the photonic lantern. The input lantern is attached to the multiplexing side of the SDM-OVNA and the output lantern to the demultiplexer. The spectrally resolved mode transfer matrix of the system is then measured for that input/output port combination and the singular value decomposition (SVD) is performed as a function of wavelength to yield the eigenvalues, which represent the losses of each of the 6 orthogonal channels the fiber supports. The ratio between the maximum and minimum eigenvalues is the mode dependent loss (MDL) and the average value is the insertion loss (IL).

 

Fig. 3 (a) SDM-OVNA system diagram. (b) Insertion Loss for 5 output ports for an example wavelength channel plan for the single-mode switching portion of the device (no photonic lanterns attached).

Download Full Size | PDF

Figure 3(b) is an example which illustrates the wavelength isolation between five of the few-mode fiber ports. In this instance the SLM mux/demux and photonic lanterns of Fig. 3(a) have been bypassed. That is, the single-mode ports of the SMF-WSS are connected directly to the delay lines of the SDM-OVNA in Fig. 3(a). This is done to provide a more accurate measurement as it removes the loss of the mode multiplexers and lanterns, increasing the dynamic range of the measurement but also ensures any measured wavelength leakage between the ports is in fact due to the WSS switching element rather than scattered light in the SDM-OVNA SLM mux/demux itself. The measured wavelength isolation is 23dB at worst. This is worse than would typically be expected for a SMF-WSS due to the fact that the input/output array has not been optimized for few-mode operation. A traditional SMF-WSS is designed to achieve high isolation when using the designated single input fiber, rather than in this case where multiple input fibers are used. The isolation could be improved if the layout and spacings of the single-mode fiber array of the SMF-WSS were optimized for few-mode operation.

The photonic lanterns and SLM mux/demux were then reattached to measure the performance of the entire FMF-WSS. Figure 4(a) illustrates the MDL of the entire system where all wavelengths have been directed to each of the 11 respective output ports. The measurement of Fig. 4(a) includes the MDL of the SLM based mode-multiplexer and demultiplexer of Fig. 3(a). The final mode transfer matrix measured (U_MEAS) is the product of three matrices given by $U_{MEAS}=U_{MUX}{U}_{DUT}{U}_{DEMUX}$. Where U_MUX is the SDM-OVNA multiplexer, U_DEMUX the demultiplexer and U_DUT is the WSS device-under-test itself. It is straightforward to measure power loss through the system to partially remove the effect of the SDM-OVNA from the results. However as the performance of the transmitter and receiver within the SDM-OVNA are described by matrices and as the modal properties of the system are sensitive to alignment and movement, it was not possible to entirely remove the MDL, and to a lesser extent the IL, of the SDM-OVNA itself from the final measurements in a meaningful way.

 

Fig. 4 (a) MDL and (b) IL for each of the 11 output ports.(c) MDL and (d) IL for various channel widths of port 1 (dark blue).

Download Full Size | PDF

Hence the values of MDL and IL shown in Figs. 4(a) and 4(b) represent an upper bound on the true values. The measurement also includes the loss associated with coupling light from the modes of a step-index fiber into the photonic lantern. The same emulated fiber profile is used at both ends and is better matched to the transmit lantern than the receive lantern with measured insertion loss of 0.9dB and 1.9dB respectively. It is likely that a splice to an actual piece of few-mode fiber would have less loss, both due to the fact that the cores could be better aligned than in the free-space optics of the SDM-OVNA but also due to the diffusion of dopants and slight tapering that occurs when cores are fusion spliced [15]. The insertion loss of the lanterns as measured by feeding light into the single-mode cores and measuring the power at the multi-mode facet on a power meter, i.e. the loss without regard for the spatial mode profile, was measured as 0.5dB and 0.8dB.

Figures 4(c) and 4(d) illustrate the spectral passband for 10, 15, 25, 50 and 100 GHz channels programmed onto port 1 (dark blue series of Figs. 4(a) and 4(b)) for both MDL and IL. It is in these two graphs that the advantage of the multi-mode to single-mode transition performed by the photonic lantern is most apparent. In contrast to a device based on an input/output few-mode fiber array [8,10,11], where the spectral roll-off of the IL and MDL are quite different for narrow channels, with MDL being the factor which ultimately limits the spectral resolution. In this implementation, the bandwidth of the IL and MDL plots are consistent, demonstrating that even though the device supports multiple spatial modes, it maintains single-mode-like performance in terms of spectral resolution (10GHz) as well as sub-GHz scale addressability and compatibility with grid-less architectures.

As mentioned above, this FMF-WSS is based on two columns of input/output single-mode fibers rather than a more traditional single-column design. An advantage of this approach is it allows some ports of the lantern to be switched independently of the others. Specifically, two single-mode lantern ports are switched together independently of the other single-mode port. This extra degree of freedom makes it easier to maintain consistent coupling between all three ports of the lanterns and hence improve MDL and IL. However as the mapping of wavelength to SLM pixel column is not exactly the same for either side of the SLM, it can come at the expense of spectral resolution. This effect is on the order of half a single pixel column of the SLM at worst, as a particular wavelength that is centered on a column on one half of the SLM, could be centered exactly between two pixel columns on the other half. When compared to measurements taken for the same components where all lanterns are attached to the same column in the single-mode fiber array of Fig. 2(a), MDL and IL are worse, but spectral resolution is superior in the single column case, with the effect only really being noticeable for the 10 GHz channel example. It should be noted that the noise floor visible in Fig. 4(c) and 4(d) is limited by the sensitivity of the SDM-OVNA rather than being a measure of the actual wavelength isolation, which was covered in Fig. 3(b).

4. Conclusion

A 1x11 MDM compatible WSS has been demonstrated with support for 3 spatial modes. We believe our device actualizes a first step to a viable optical switching solution for MDM transmission networks. The device can readily be extended to accommodate more fiber modes or output ports.