Channel measurement receivers are investigated in the context of uncoded digital communication through the clear turbulent atmosphere at optical frequencies. Such receivers exploit both the long (few millisecond) coherence times and the narrow spatial coherence of the turbulence-induced fading process in the channel to improve the reliability of communication. The channel measurement problem is formulated in terms of hypothesis testing so that the receivers and their performance nay be determined from the principles of statistical decision theory. These principles are used to discover how the decision process can best make use of data from the N most recent past signalling bauds in improving the reliability of the current decision.nThe channel fading process is assumed to be lognormal;many results for Rayleigh fading are included as well. Two optical front end processors are considered-heterodyne and direct detection. The appropriate output process models for these receivers are conditionally Gaussian and doubly stochastic Poisson, respectively.nPerfect measurement bounds are derived to ascertain the extent to which channel measurement can improve performance. Measurement receivers are found which are optimum in the sense that no other receiver having access to the sane data base can have superior performance. Complexities in both the receiver structures and performance calculations lead us to consider alternative processors. Among these are schemes which incorporate past decisions into the decision process, and others which execute diversity combining in a majority rule sense.nWe show that for heterodyne receivers an increase of at least 3 dB in effective signal to noise ratio can be gained by measurement;receivers are exhibited whose performance approaches the perfect measurement bound as N increases. For direct detection lesser gains are predicted. We are unable to demonstrate that our measurement, receivers realize the perfect measurement potential.
展开▼
