Dependence of error rate on signal-to-noise ratio in fiber-optic communication systems with phase-induced intensity noise

It is experimentally demonstrated that when multipath optical-fiber systems are driven by single-mode sources whose linewidth is small compared with the detection system's electrical bandwidth, the amplitude distribution of the phase-induced intensity noise (PIIN) can become highly nonGaussian with extremely short tails. Consequently, a given low error rate may be maintained at significantly smaller signal-to-noise (S/N) ratio values than Gaussian statistics would predict. This effect is particularly prominent at the large S/N values at which digital communication systems usually operate. The theory of PIIN is briefly described, including a quantitative treatment of the photocurrent S/N bit-error-rate (BER) relationship in systems where both PIIN and thermal noise exist. A technique that enables the measurement of the S/N-BER relationship well into the tails of the noise-photocurrent distribution, where ordinary histogramming techniques fail due to insufficient resolving power, is also described. >