Adaptive Filter Modeling for Communication System Interference Rejection

The determination of communication system performance is necessarily constrained by received "front-end" signal-to-noise ratio (SNK), among other limiting system parameters. Currently, there exist two dual approaches for enhancing the system "front-end" SNR at the antenna output: adaptive array processing, in which the interference is reduced by steering a null in the direction of the jammer, and adaptive noise cancelling (ANC), in which a null is placed in the direction of the signal source in order to get an independent estimate for the interfering noise field. Traditionally, ANC\´s have employed tapped delay line (TDL) implementations and in this paper a rationale is provided for using an infinite impulse response (IIR) configuration in order to improve rejection of broadband jamming signals. It is also demonstrated that nominal finite impulse response (FIR) representations are designed to work quite successfully when the jamming source is intrinsically narrowband. Central to this discussion is the notion of moving average (MA) versus autoregressive (AR) filters used as estimators, for determining the nature of the process to be modeled. The approach in this paper is to discuss the filter estimation problem from the view point of examining the spectral characteristics of the interference process. It is demonstrated that the response of the AR estimator is quite good against broadband signals, when compared with the normal MA estimator. The "peaky" performance of the AR estimator bcomes quite "smooth" when sufficient numbers of data points are included for the broadband process.