Convolutional Code Performance Analysis of Jammed Spread Spectrum Channels Without Side Information.

The goal of this dissertation is to analyze systems where soft decision decoding may be used in the presence of worst case pulse jamming with an unknown jammer state. Previous work has shown that a decoder, using soft decision maximum likelihood decoding, must know, for each chip, whether that chip is jammed or not. Otherwise it gives very poor performance. The simplest solution to the problem is shown to be clamping the level of the signal at the input of the decoder. This allows the use of soft decisions in the decoder without paying the price in poor performance with worst case pulse jamming. It was shown that this system does better than a hard decision decoder and may be improved upon even more by controlling the clamping level as a function of the jammer's duty cycle and signal -to-noise ratio. An explanation is given for the the poor performance of the soft decision receiver with unknown jammer states and for the good performance of the clamped receiver with unknown jammer states. Systems that try to estimate the state of the jammer and use soft decision decoding are also investigated. The performance of a receiver which produces an estimate of the state of the channel for each received chip, and chooses its metrics accordingly, is analyzed as a function of the reliability of the estimate. Two methods of estimating the jammer state are considered. The first method bases its estimate on the amplitude of the received signal. The second method assumes an external estimate of the jammer state. It was shown that basing the estimates on the signal level is necessary for good performance if there is some probability that jammed chips would be incorrectly estimated as unjammed by the external estimator. These systems are investigated for both the direct sequence and the frequency hopped spread spectrum channel. The same sort of solutions that work for direct sequence also work for frequency hopped systems. The weakness of systems that use an external estimate on the state of the jammer, independent of the signal, is shown to be even more apparent in frequency hopped systems. Finally, error probabilities are estimated for actual codes to confirm the results obtained from the random coding bounds.