In digital communication systems, several types of impairments may be introduced to the signal. These impairments result in degraded system performance; for example, high bit-error-rate or power penalty. For optical communication systems, in this thesis, these impairments are categorized into four types; that is, thermal noise, shot noise, signal-dependent noise, and intersymbol interference (ISI). By using a Gaussian approximation, effects of the first three impairments are analyzed. It is shown that signal-dependent noise introduces an error floor to the system and the bit-error-rate is considerably degraded if a nonzero-extinction ratio is applied to the system. It is shown that if the decision threshold at the decision circuit is set improperly, more received power is required to keep the bit-error-rate constant.

Three main components in the system (i.e., transmitter, optical fiber, and receiver) are modeled as Butterworth filters. ISI from this model is determined by computer simulation. A high ISI is from a small system bandwidth. It is shown that a minimum power penalty can be achieved if the transmitter and receiver bandwidths are matched and fixed, and the ratio of fiber bandwidth to bit rate is 0.85. Comparing ISI from this model to ISI from raised cosine- rolloff filters, it is shown that at some particular bandwidths ISI from raised cosine-rolloff filters is much lower that that from this model. However, if the transmitter and receiver bandwidths are not matched and are not equal to these bandwidths, ISI from this model is lower than ISI from raised cosine-rolloff filters.