Investigation of the combined effect of polarization-mode dispersion and polarization-dependent loss on system performance

Polarization effects such as polarization-mode dispersion (PMD) and polarization-dependent loss (PDL) cause a variety of impairments to optical fiber communication systems. PMD induces pulse distortions and depolarization of the signal that degrade system performance. PDL can cause the initially unpolarized noise become partially polarized and thus increases the system performance variation. When combined together, they can produce surprising effects that cannot be predicted by their effects acting alone. To investigate their combined effect on system performance it is necessary to develop a receiver model that can handle a depolarized signal due to PMD and partially polarized noise due to PDL. In this dissertation, my colleagues and I derived a formula for the variance of the received electric current of a depolarized signal due to first-order PMD and partially polarized noise. From this formula and the formulae for the mean of the received electric current, we were able to calculate the system performance, e.g. the Q-factor due to the combined effect of PMD and PDL. We validated the formula by comparing Q-factors obtained from the theory, Monte Carlo simulations and experiments and found good agreements. We also found that the partially polarized noise can induce large system performance variation when the signal is depolarized. This finding is the same as when the signal is polarized. However, the beating between the signal and the noise is totally different when the signal is polarized or depolarized. In systems where the signal is depolarized, the angle between the SOP of the signal and the polarized part of the noise varies as a function of time within each bit window, while the angle is constant in systems where the signal is polarized. As a result, the beating between the signal and the noise varies over time within each bit window when the signal is depolarized and this can cause complications in determining the clock time.; In addition to the validation of the model, we also investigated the sensitivity of the Q-factor to the inherent measurement uncertainty of the pulse width of the signal, the OSNR, and the receiver filter bandwidths. This study is especially important to experimentalists and system designers to estimate the level of uncertainty of system performance in practical systems. We found that each parameter induced a variation of the Q-factor from 0.1 to 0.3 out of a mean Q-factor of 7.3 in linear scale when those parameters were varied in reasonable ranges. Our model can be used in On-Off-Keyed systems where first-order PMD and PDL cause significant penalties.