Near real-time adaptive radar processing using analog dithers

This paper explores the performance of a novel analog technique for estimating the gradient of a cost function and its application to adaptive processing in radar. The technique relies on using analog sinusoidal dithers to estimate components of the gradient vector in near real-time with a latency on the order of microseconds. The advantages of this approach are that the adaptive processor can quickly adapt to nonstationary environments and the dynamic range at the input to the adaptive processor is not constrained by the limits of an analog-to-digital converter. Simulated results are shown for an adaptive sidelobe canceler in a radar configured with a main beam sum channel and auxiliary channels. Some drawbacks of the proposed technique for sidelobe cancellation include that it is suitable only for steepest descent type algorithms which have an inherently slower convergence rate, as well as the requirement to carefully account for or remove any time delays, and gain or phase imbalances in the analog dither circuitry, so as not to degrade the estimate of the gradient.