Broadband Integrated Fabry-Perot Electro-Optic Switch M. Ménard and A. G. Kirk Department of Electrical and Computer Engineering, McGill University 3480 University Street, Montréal, QC H3A 2A7 Canada Abstract: Electro-optic switches based on tunable Fabry-Perot filters suffer from limited transmission bandwidth. This limitation can be overcome through the use of a comb filter and we demonstrate this concept in a photonic integrated circuit. Keywords: Electro-optic Switch, Integrated Optics, Comb Filter Introduction With the advent of reconfigurable optical add/drop multiplexers (ROADMs), wavelength routing is gradually being integrated into optical telecommunication networks. However, current ROADMs use slow switching technologies [1], such as micro-opto-electromechanical systems or thermo-optic switching. To build truly agile optical networks, where wavelength connections between nodes can be established in real time, the next generation of routing systems will require sub-microsecond switching times [2]. The electro-optical (EO) effect permits switching within this time frame [3] and is already used in modulators. Over the past 40 years, multiple device configurations have been investigated to fabricate EO switches, both in free-space and guided wave systems [4-7]. However, scalability is still an issue, as free-space systems are large [7] and guided wave systems require complex interconnection networks [5] or complex fabrication [6]. In this paper, we present an optical space switch based on tunable coupled-cavity Fabry-Perot (FP) filters. It is implemented in a slab waveguide, which enables us to take advantage of the benefits of both free-space and guided wave propagation. Furthermore, it can be used as a building block for both crossbar and Banyan switch fabrics. 2 Switch Design Although the refractive index change induced by the EO effect can be rapidly modulated [3], on the order of nanoseconds, it is also very small (~10 -3 in III-V). Thus, the first challenge when designing a switching element based on this phenomenon is to find a way to exploit it efficiently. Introducing resonance into the light path with FP cavities amplifies the effect of the refractive index change and creates the equivalent of mirrors that can be turned on or off at GHz frequencies. However, there are numerous aspects that require careful engineering in order to obtain a practical device. The small refractive index change provided by the EO effect can only shift the transmission peak of a first-order FP filter by about 1 nm around 1550 nm in GaAs. In order to increase the spectral bandwidth, we have decreased the free-spectral range of the FP filter to create a comb filter that transmits every other wavelength of the channel grid, as illustrated in fig. 1. Thus, it is now possible to switch the filter response from reflection to transmission for any channel over a 30 nm bandwidth with only the EO effect. A drawback of doing this is that the filter looses its demultiplexing capability since it now transmits all “odd” or “even” channels. However, a colorless switch can be obtained by using a static interleaver to separate the “odd” and “even” bands and feed them to two different switches, whose outputs are then recombined by a second interleaver. Furthermore, “odd” and “even” channels can be used as wavelength bands to provide different levels of granularity. Fig. 1: Illustration of the tuneable comb filter operation. FP filters working at oblique incidence are frequently used in free-space systems where they are often implemented as thin-film filters. In photonic integrated circuits, refractive index modulations can be obtained through etching different depths in a slab waveguide. The maximum index contrast is achieved when the waveguide is completely etched through. This approach is used to define the cavity mirrors. Furthermore, since light beams in integrated optics have cross-sections that are much smaller than in free-space, care must be taken to ensure that the light beam angular spectrum can be transmitted and reflected undistorted by the filter. This imposes a minimum size on the beam width, and thus on the device, which is a function of the incident angle on the filter. Calculations combining plane wave decomposition of the fundamental guided mode and transfer matrix analysis showed that 99.99% of the incident beam angular spectrum must be within the 90% filter passband to preserve its plane wave spectral response. As the angle of incidence is increased, the filter angular spectrum quickly decreases. For instance, a filter with a 0.6 nm 3dB passband at 1550 nm requires a beam waist radius larger than 237.5 μm at 10 degrees incidence. Similarly to thin-film devices, the filter response can be tailored by adjusting the number of cavities and their thickness. A flat response with a 3dB passband of 0.6 nm is