Linear Direct-Detection Fiber-Optic Receiver Optimization in the Presence of Intersymbol Interference

A technique is developed for optimizing communication receivers by minimizing a Chernoff bound (CB) on the Bayes cost function. Being couched in terms of moment generating functions, the procedure is well suited to optical systems. Optimal CB receivers are derived and shown to coincide exactly with known minimum cost Bayes receivers for certain special cases of both electrical and optical communications for which closed form analytic solutions are possible. This provides endorsement for the method, which is then applied to more complex practical cases of linear optical fiber receivers in which both colored Gaussian noise and intersymbol interference play a part. The CB receivers are shown to offer some improvement compared with a zero forcing raised cosine equalizing receiver such as is commonly adopted, and all the evidence suggests that CB receivers achieve performance very close to the best possible with a linear receiver.