All-optical polarization control and noise cleaning based on a nonlinear lossless polarizer

We propose an all-optical ber-based device able to accomplish both polarization control and OSNR enhancement of an amplitude modulated optical signal, a ected by unpolarized additive white Gaussian noise, at the same time. The proposed noise cleaning device is made of a nonlinear lossless polarizer (NLP), that performs polarization control, followed by an ideal polarizing lter, that removes the orthogonally polarized half of additive noise. The NLP transforms every input signal polarization into a unique, well de ned output polarization (without any loss of signal energy) and its task is to impose a signal polarization aligned with the transparent eigenstate of the polarizing lter. In order to e ectively control the polarization of the modulated signal, we show that two di erent NLP con gurations (with counter- or co-propagating pump laser) are needed, as a function of the signal polarization coherence time. The NLP is designed so that polarization attraction is e ective only on the "noiseless" (i.e., information-bearing) component of the signal and not on noise, that remains unpolarized at the NLP output. Hence, the proposed device is able to discriminate signal power (that is preserved) from in-band noise power (that is partly suppressed). Since signal repolarization is detrimental if applied to polarization-multiplexed formats, the noise cleaner application is limited here to "legacy" links, with 10 Gb/s OOK modulation, still representing the most common format in deployed networks. By employing the appropriate NLP con gurations, we obtain an OSNR gain close to 3dB. Furthermore, we show how the achievable OSNR gain can be estimated theoretically.