Abstract
In this paper, we study the validity and limitations of the additive white
Gaussian noise (AWGN) model in quasi-linear, long-haul, return-to-zero, direct-detection
optical fiber communications systems. Our approach is to compare bit-error ratios (BERs)
computed using the additive white Gaussian noise method to those obtained using standard
and multicanonical Monte Carlo (MMC) simulations and to a noise-linearization method,
referred to as the noise covariance matrix (NCM) method. We show that the AWGN method
provides a very good approximation to the actual system BER for power levels and
dispersion profiles that are used in typical modern-day quasi-linear systems. For
example, the BER obtained using the AWGN method is within a factor of 4 of the actual
system BER computed using MMC simulations for a realistic 10 Gb/s, 6000 km system based
on dispersion-shifted fiber in which the peak signal power at the transmitter is 1 mW
and the absolute residual dispersion at the receiver is less than 200ps/nm. However,
when the peak power is increased to about 4 mW, or the average map dispersion is zero
and the absolute residual dispersion exceeds 200 ps/nm, the AWGN and NCM methods may
simultaneously breakdown due to a combination of nonlinear signal–noise and noise–noise
interactions during transmission. In addition, for a 5000 km system based on
low-nonlinearity $D_+$ and $D_-$ fiber with an average map dispersion that is 4% of the
dispersion variation within the map, and that operates at a peak power of 5 mW, we find
that the BERs obtained using the AWGN and NCM methods are about 500 times smaller than
the actual system BER computed using MMC simulations.
© 2009 IEEE
PDF Article
More Like This
Cited By
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
Contact your librarian or system administrator
or
Login to access Optica Member Subscription