|Nulling Over Extremely Wide Bandwidths
When Using Stretch Processing
Richard M. Davis, Jose A. Torres,
J. David R. Kramer, and Ronald L. Fante
The MITRE Corporation
202 Burlington Road
Bedford, MA 01730-1420
tel: (781) 271-5501
Abstract This paper describes a technique for performing adaptive nulling over extremely wide bandwidths. The technique is applicable to radar systems that use linear frequency modulated (LFM) waveforms and stretch processing. It is assumed that the range uncertainty of the target is a small percentage of the equivalent range extent of the uncompressed pulse. The assumptions allow cancellation to take place in the time domain in a narrowband sliding filter that keeps up with the chirp rate or in the frequency domain. The new approach supports nulling performance over Gigahertz of bandwidth comparable to that achieved over about 10 Megahertz using the same number of degrees of freedom.
Traditional approaches to nulling over wide bandwidths utilize adaptive finite impulse response (FIR) filters or subbanding techniques. Both techniques require large numbers of adaptive weights. If M time taps per filter are required to null a single jammer, at least J FIR filters and JM weights will be required to null J jammers. Subbanding at RF or IF requires many receiver chains, while subbanding digitally requires analog-to-digital conversion of the entire cancellation bandwidth.
The author's approach assumes that the wideband mode will be used only when the target's range is known to within a small fraction of the range extent of the uncompressed pulse. The assumption allows us to destretch the pulse and then perform the nulling in a narrowband filter in the time domain following the destretch operation, but before the final FFT. Alternatively nulling can be performed in the frequency domain after the final FFT. The latter approach requires the use of out-of-band correlation. Many algorithms exist for calculating the adaptive weights.
The authors will present a theoretical result showing that the same weights that null
one frequency bin can be used to null all bins. Computer generated numerical results will
also be presented. Both time and frequency domain performance will be shown using the
sample matrix inversion (SMI) algorithm to determine the adaptive weights.
Direct comments and questions to: firstname.lastname@example.org
© MIT Lincoln Laboratory. All rights reserved.