An adaptive technique is presented for suppressing clutter in a moving target indicator radar signal processor. The technique is based on the assumption that radar clutter can be modeled as a Mth-order auto-regressive (AR) process and employs a lattice prediction error filter structure to suppress clutter without significantly distorting the target signal. Filter coefficients are the reflection coefficients determined using one of the three variations of Burg's AR parameter estimation algorithm due to Andersen, Makhoul and Marple. These filter designs are applied to second-order Butterworth-shaped spectra representing both ground and storm clutter, and their performances are compared with each other and with more conventional methods in terms of improvement in signal-to-clutter ratio between filter input and output. Results are presented which show the dependence upon such factors as sample size, filter order, and doppler frequency.

An algorithm is presented for designing multi-band FIR digital filters to suppress wideband or bimodal clutter in MTI radars. The technique is an iterative process in which the passband edges are varied to maximize the usable bandwidth over which a signal is detected with acceptable probability while maintaining specified clutter rejection. The clutter-bandwidth tradeoff is shown to be a function of such interrelated filter parameters as number of taps, transition widths, passband ripple, variable stopband attenuation, and stopband width. A 25-tap bandstop filter design is presented which provides 70% usable bandwidth while maintaining 20 dB wideband clutter rejection. The design is then modified to include an additional 50 dB ground clutter attenuation. The resulting 26-tap multi-band filter meets all specifications while still retaining 60% usable bandwidth. An alternative technique for bimodal clutter rejection is presented which utilizes the natural null at dc for an N-even bandstop filter to attenuate the ground component.

A technique is presented for improving the bandwidth throughout which a radar return can be detected while keeping the ground-clutter power rejection below an acceptable level. The processor employs a high-pass digital filter which has been designed in accordance with the McClellan-Parks algorithm with the pass-band lower edge selected in an optimum fashion when the other filter design parameters are specified. Conventional approaches to the design of a moving-target indicator (MTI) filter such as the three-pulse canceller and the covariance design method typically improve the signal-to-noise ratio over less than 50% of the range between dc and the pulse-repetition frequency whereas this technique extends the usable bandwidth to 80% or more. A design example is included which utilizes parameter values from a C-band radar and demonstrates the interactive influence of such filter parameters as the number of weights, passband ripple and bandedge, and stopband attenuation and cut-off.

A general performance equation is derived for digital receivers in the presence of impulse noise. Two bounding techniques are employed to simplify the results. First, the Schwarz inequality removes the dependence on both the impulse noise waveforms and the signal-set basis functions. Second, the Chebyshev inequality is applied to the Schwarz results to additionally eliminate the statistical dependence on the probability density function of the noise waveform weighting factor. The bound curves are compared to specific theoretical results where the noise waveforms and probability density functions have been assumed. The tightness of the bound curves is a function of the signal-energy-to-noise-weighting-factor variance ratio and converges for the Schwarz bound as the ratio increases. Considerably looser bounds are obtained by the Chebyshev technique, but rough estimates of system performance can be obtained when the signalenergy-to-signal-energy-to-noise-variance ratio is the only known parameter.

A time-domain technique is presented for the design of frequency-sampling and transversal digital filters for use in equalizing channels with known impulse responses. The technique uses a linearprogramming algorithm to specify filter multiplier coefficients that satisfy constraint equations based upon specified values and derivatives of the desired output signal. It is shown that the transversal equalizer design requires fewer multipliers and performs in a superior fashion in the presence of timing jitter. The effects of changing the number and type of constraints upon such parameters as equalizer complexity, transmitted energy, and residual distortion are presented.

The percentage of total power passed through ideal bandlimited channels with bandwidths equal to various multiples of the bit rate is examined for 14 unique binary baseband signaling schemes. The channel bandwidth required to pass 5 percent of the total power is determined as a measure of the need for near-dc channel response. Conclusions are reached as to which signals would be most suitable for high-rate magnetic tape recording.