EVALUATION AND COMPARISON OF METHODS OF FILTERING

The effectiveness of various electrical filters in simultaneous smoothing and differentiating azimuth-angle data was analyzed to determine the optimum method of eliminating noise, N(t), superimposed on the signal, A(t). Two types of signals were used; one was associated with a target on a straight-line path at an angle beta from the ownership course, and the other was associated with the target on a pure pursuit course (parabolic path) relative to the ownship. The noise function was assumed to have a typical form of the autocorrelation function R(t) = sigma squared eta sub minus a/tl, where sigma squared = R (theta) gives the mean square value of the noise amplitude N(t). Evaluation criteria based on the resulting signal distortion and the value of N(t) in the output of the filter were selected and brief derivation is included of the Zadeh-Ragazzini (Z-R) theoretical expressions for filtering (J. Appl. Phys. 21 7:645-655, 1950) which are based upon the assumption that the signal is representable as a polynomial P(t) of degree not higher than a specified number n. The comparative evaluation of various RC filters on the basis of the Z-R theory and the criteria indicated an optimum filter network transfer function which resulted in a mean square noise output less than 25 sq yd/sec/sec for a 37-ft-long attacking plane. An experimental approach was suggested for future investigations.