On the Gaussianly-distributed Q-factor in optical communication systems

We show that the average-Q method may bring significant error on BER if the Q variations are mainly caused by system performance fluctuations. A new Q-function with an effective Q-factor was derived and evaluated by assuming measured Q values followed a Gaussian distribution. Results showed that traditional Q-function with average Q values underestimate the system BER. In the system design point of view, this implies that margins need to be left on the measured average Q to achieve specific bit error-rates.