On the node ordering of progressive polynomial approximation for the sensor linearization

Many sensors exhibit nonlinear dependence between their input and output variables and specific techniques are often applied for the linearization of their transfer characteristics. Some of them include additional analog circuits, while the others are based on different numerical procedures. One commonly used software solution is Progressive Polynomial Approximation. This method for sensor transfer function linearization shows strong dependence on the order of selected nodes in the linearization vector. There are several modifications of this method which enhance its effectiveness but require extensive computational time. This paper proposes the methodology that shows improvement over Progressive Polynomial Approximation without additional increase of complexity. It concerns the order of linearization nodes in linearization vector. The optimal order of nodes is determined on the basis of sensor transfer function concavity. The proposed methodology is compared to the previously reported methods on a set of analytical functions. It is then implemented in the temperature measurement system using a set of thermistors with negative temperature coefficients. It is shown that its implementation in the low-cost microcontrollers integrated into the nodes of reconfigurable sensor networks is justified.