Implementing a matrix-inversion algorithm in a limited-precision adaptive antenna array processor / by Robert W. Jenkins. : Co24-3/8-725-1984E-PDF

"The effect of computation accuracy on the performance of a small adaptive antenna system using a digitally-implemented matrix inversion algorithm is examined. A single-board array processor with 16-bit mantissa floating-point accuracy is compared with a VAX-11/750 computer having 24-bit accuracy in the single-precision mode, and 56-bit accuracy in the double-precision mode. The effects of limited precision on array performance are worse when the signal covariance matrix to be inverted is ill-conditioned (i.e., nearly singular), as is the case with an n-element array in the presence of less than n unrelated signals and a low system noise level. Using a simulated signal environment consisting of a wanted communications signal, a single jamming signal, and a variable system noise level, the performance of the array is evaluated. A need for about -20 dB of artificially-injected system noise (relative to the total signal) is demonstrated in the case of the singleboard array processor, in order that the covariance matrix be sufficiently well-conditioned for adequate array performance"--Abstract, page 1.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.892625&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs