Since the invention of fiber optic communications, researchers have sought to increase the amount of information a fiber can carry, by encoding data channels in wavelength, polarization, phase, and amplitude. Now space is the next frontier in advanced fiber networks. In space division multiplexing (SDM, also known as mode division multiplexing or MDM), different data channels are encoded in different spatial modes, analogous to wavelength coding in wavelength division multiplexing (WDM). Which of course implies that the transmission fiber in MDM is multimoded. This represents a radical break from traditional optical networks based on single mode fibers (SMF), as it will require innovations on nearly every level -- in hardware alone, one needs new transmission fibers, new amplifiers, new dispersion compensators, new routers and couplers, and new understanding of nonlinear impairments.

MDM is still in its early days, and there are currently two visions for how it might be implemented. SMF is ubiquitous in part because it allows for precise control of pulse propagation, as only one mode is guided with well-defined group index and dispersion. In multimode fiber (MMF), guided modes generally mix due to bends, stress, or weak longitudinal variations in the fiber, so that even with a pure single mode excitation, the output is partitioned amongst all modes. So then how can SDM work? One method is to take a page from wireless networking and exploit mode mixing. In MIMO (multiple in/multiple out), modal dispersion actually increases channel capacity. The signals need to be demultiplexed electronically with coherent detection, so MIMO requires additional power consumption and is limited in the maximum bit rate per SDM channel. Mode mixing can be inhibited by designing a few mode fiber such that the modes are well-separated in effective index. In this version of SDM, distinct spatial modes can be launched, transmitted and directly detected. So how to mux/demux the modes? Ideally, one would want a lossless, wavelength-independent integrated device which takes a single mode input that can be interfaced with existing emitters and detectors, avoiding the need for free-space beam shaping or steering. Yerolatsitis and co-authors at the University of Bath have recently demonstrated a path to such a device, based on an all-fiber null coupler design.

In a fiber null-coupler, multiple fiber inputs are adiabatically fused and tapered together to form a common multimode core at the waist. The adiabaticity requirement ensures that light launched into a pure modal state remains in a pure modal state as it propagates. This means that if the input waveguides are dissimilar, they can be ordered with distinct propagation constants, and that ordering is maintained when the light reaches the multimode waist: light launched to the input port with the n^th largest propagation constant evolves into the n^th mode of the multimode waist. Most importantly, the operating principle of the device has no explicit wavelength dependence, so the bandwidth for mode conversion is very broad. The authors demonstrate this experimentally with two embodiments, both of which generate three distinct modes. They use selective hole collapse and tapering to make a three-core photonic crystal fiber (PCF) evolve into a multimode core PCF with form birefringence, which lifts the degeneracy between the even and odd LP_1,1 states. In the second embodiment, which better resembles the ideal device described above, they taper and fuse three single mode fibers in a fluorine glass capillary jacket, again generating form birefringence at the multimode waist to make each mode distinct. In both cases, they achieve > 15 dB mode purity and losses < 0.6 dB.

Though these results are promising, one important question is whether the null coupler can be scaled past three modes. The authors show that, in order to maintain adiabaticity, the device length scales with the square of the number of modes. However, all is not lost: the length requirement can be relaxed if the core/cladding ratio at the multimode waist becomes large, and a 12 mode SDM multiplexer seems entirely feasible. And so on to making 12 mode EDFAs…

---

---

 Add Comment

You must log in to add comments.