Canceler Performance Degradation Due to Estimation Noise

The multiple sidelobe canceler (MSLC) configured using the gradient descent algorithm converges to an approximation to the optimum weight vector. Since the weight vector estimate is obtained from a finite data set, there is a random error that increases the steady state residue power. An exact analysis of the steady state residue power for a digital implementation is performed. The analysis predicts that as the algorithm's gain (step size) is increased, the steady state residue power tends towards infinity. The gain that causes infinite power is smaller than the gain that causes instability for the average weight vector analysis. It is shown that the steady state residue power is equal to the minimum residue power multiplied by a factor that depends upon ratios of the eigenvalues. The steady state residue power is evaluated for some special cases. It is shown that the value of the factor is greatest when the interference is due to a single jammer.