Control Sensor Linearization Using Artificial Neural Networks

Traditionally, the issues of cost, size, and weight of artificial neural network implementations have not been the primary concern of researchers. These issues are important in many applications such as those required in space travel, high-volume commercial products, or products with size limitations. In this paper, we discuss methods to improve the characteristics of control sensors using compact and low-cost circuitry. Our objective is to extend the linear region of operation of a nonlinear sensor using artificial neural networks. An analog circuit approach was investigated for high-speed applications and a microcontroller approach for low-speed applications. The methods are applied to the design of a discrete-component phase-locked loop. Both approaches resulted in doubling the sensor's linear range.