IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 15, NO. 1, JANUARY 2003 141 A Novel Dynamic Crosstalk Characterization Technique for 3-D Photonic Crossconnects Cedric F. Lam, Senior Member, IEEE, Misha Boroditsky, Senior Member, IEEE, Bhavesh Desai, and Nicholas J. Frigo, Member, IEEE Abstract—We introduce a novel sensitive technique to charac- terize dynamic crosstalk in three-dimensional photonic crosscon- nects (PXCs). This nonintrusive technique enables end users to evaluate PXC dynamic performances without special access to the internal photonic-switching fabric. Our results confirm that while dynamic crosstalk can be significantly higher than static crosstalk, it has little system performance impact under normal operating conditions. We discuss methods to mitigate the effects of dynamic crosstalk at the end of this letter. Index Terms—Bit-error rate (BER), crosstalk characterization, crosstalk mitigation, dynamic crosstalk, photonic crossconnect. I. INTRODUCTION T HE PHOTONIC crossconnect (PXC) has been regarded as a revolutionary technology for next-generation optical networks. It has the advantage of being protocol agnostic and bit rate independent. PXCs based on three-dimensional (3-D) micro-electro-mechanical systems (MEMS) are promising be- cause of their ability to scale to large port counts linearly [1]. The 3-D MEMS PXCs usually contain an array of input mirrors and an array of output mirrors, both of which are arranged on a two-dimensional lattice of square, rectangular, or triangular patterns, etc. An example is shown in Fig. 1(a). Fig. 1(b) illustrates the operating principle of 3-D MEMS PXCs. Connections are established between input and output port optical collimators by free-space propagation of the optical signal and proper tilting of an input and an output mirror. Only linear strings of input and output mirrors are shown in Fig. 1(b). The solid lines in Fig. 1(b) show five adjacent parallel beams forming optical connections. Suppose that the beam from input port 4 [in Fig. 1(b)] is now redirected from output port 4 to output port 2 (the original beam aiming at output port 2 from input port 2 is first redirected to somewhere else) to make a new connection as shown by the dotted lines. To make this connection, mirrors I4 and O2 must move. During the transition, we would expect that, depending on the order and manner of the mirror motions, light from input port 4 could unintentionally and briefly enter output port 3. This would lead to channel crosstalk of a highly dynamic nature. The magnitude of this crosstalk depends on the optical coupling properties of the port collimators and the details of how light is scattered Manuscript received May 30, 2002; revised September 18, 2002. C. F. Lam was with AT&T Labs-Research, Middletown, NJ 07748 USA. He is now with Opvista Inc., Irvine, CA 92604 USA. (e-mail: cflam@ieee.org). M. Boroditsky, B. Desai, and N. J. Frigo are with AT&T Labs-Research, Middletown, NJ 07748 USA. Digital Object Identifier 10.1109/LPT.2002.805775 (a) (b) Fig. 1. (a) An example input and output mirror arrays used in 3-D MEMS PXCs. The mirrors are arranged as a triangular lattice in this example. (b) Operating principle of 3-D MEMS PXCs. from the structures between mirrors. The severity of this effect is greatest when adjacent ports make parallel connections, as in Fig. 1(b). Impairments of this nature could potentially cause error bursts, which, if serious enough, could further trigger other network activities, such as alarm signals and protection switching. PXCs switch optical beams on millisecond time scales. To the best of our knowledge, without the ability to hold the mirrors in intermediate positions, there is no standard method to capture and study the above mentioned dynamic crosstalk effects at the system level. We present a novel technique to measure this dy- namic crosstalk in PXCs. II. EXPERIMENTAL SETUP AND RESULTS Preproduction versions of PXCs from several vendors were tested in our experiment, using a procedure which is essen- 1041-1135/03$17.00 © 2003 IEEE