I. Introduction
Routing and monitoring of today's optical networks are usually done in the electrical domain by means of electrical transponders using, e.g., the SONET/SDH protocols. However, electrical transponders for high bit rates (40 Gb/s) will be very expensive, and the number of transponders can be reduced by using all-optical equipment. Since the O/E/O conversions then are not performed, the monitoring opportunity in the electrical domain vanishes. Still, performance monitoring is required and a solution is to introduce optical performance monitoring in the network. Optical performance monitoring can be accomplished by measuring the optical signal-to-noise ratio (OSNR) of each channel using spectrum analysis. The OSNR measurement is conventionally done by assuming that it is possible to interpolate the background noise level within a channel from the noise levels beside the channel [1]. However, in future optical networks this will not always be possible, e.g., when all-optical equipment is reshaping the background noise, or when the noise levels between the channels cannot be observed when the channel spacing is dense. These problems can be overcome by adding polarization analysis to conventional spectrum analysis, for example, by measuring the Stokes parameters from which an additional OSNR measurement can be obtained. Furthermore, polarization analysis of the network makes it possible to perform polarization-mode dispersion (PMD) monitoring. This paper evaluates the performance of OSNR monitoring using polarization analysis by measuring the Stokes parameters. The measurement method is applied in a realistic wavelength-division multiplexing (WDM) system setup and is compared to conventional OSNR measurements.