Experimental test of maximum likelihood thresholds based on Kalman filter estimates in on–off keyed laser communications in atmospheric turbulence

A new approach for mitigating atmospheric turbulence effects on free-space laser communication performance is presented. The method is based on evaluating Maximum Likelihood Thresholds using Kalman filter estimates in on-off keyed laser communications in atmospheric turbulence. Experimental results presented in this chapter clearly demonstrate the usefulness of the method. Although the results are shown for low data rate of 6.3 Kbps, the concept is valid for much higher data rate taking advantage of the higher bandwidth capability of laser communications. When looking at the table of the comparison of various threshold approaches, it is shown clearly that the anticipatory a posteriori processing of 50 data points provides a very good bit error performance, if the extra computational burden of processing 50 prior data points for every likelihood ratio test at every sample point is acceptable. The next best approach is the use of the dual Kalman filters which yields a BER that is within a factor of two of a posteriori approach, but without having to reprocess 50 prior data points before each threshold decision. Clearly, this is an acceptable alternative to the a posteriori approach with its more extensive processing and its unavoidable delays. Because the required update rates of h is determined by the slowly varying atmospheric turbulence, high data rate links using the dual Kalman approach will allow many more times the number of data samples and bits to be tested against and validated by the most up-to-date h. The dual Kalman approach is a new approach that is expected to give better BER performance in low and high data rate atmospheric optical links and to simplify the necessary processing in the receiver. The central issue addressed in this chapter is whether Kalman Filtering is helpful in dealing with atmospheric turbulence, and this chapter shows that it is within the bounds of statistical approximations.