Search

The optimum structure of a signal to be transmitted over a randomly time-varying and frequency-selective medium is investigated. A model is developed that treats the medium as a randomly time-varying linear filter. By viewing the filter's transfer function as a homogeneous random field on the time-frequency plane, a second-order theory results that relates various second-order measures of the time and frequency structures of input and output processes. A Neyman-Pearson detector is assumed, and a signal-design strategy, based on the asymptotic behavior of the false-dismissal probability when the detector is presented with a sequence of observations of the medium output, is developed. This approach leads to the strategy of maximizing the Kullback-Leibler information number. It is shown that this criterion minimizes the false-dismissal probability for any reasonable false-alarm probability when the medium satisfies Price's 'low-energy coherence' condition.

The irreducible binary error probability of an FSK system with memory operating over a dispersive channel is analyzed. Binary error probability is obtained as a function of both time-delay and Doppler spread for a simple HF channel model. In addition, the effectiveness of correlated diversity receptions in combatting channel dispersion is investigated.

The use of HF communication channels has been minimized as other means of communication (cable, satellite) have become available. To improve the quality and use of these channels, and also to provide optimum channels for those users who are exclusively dependent upon HF communication, the Common User Radio Transmission Sounding (CURTS) system was originated. The CURTS Phase II Automatic Frequency Selection system is described. Detailed descriptions are given of the digital signal processing, the signal energy and time-delay spread prediction, and the frequency quality and ordering techniques. An evaluation is made of both the signal processing and the operational effectiveness of the system. Performance based on six months of data on 11 DCS trunks terminating at Wheeler AFB and McClellan AFB indicate the following: Circuit outage is reduced 20-45 percent; Trunk availability is increased 45-65 percent; Binary error rates of 10 to the minus 5th power or better can be maintained for 90 percent of the time.