A flexible tool for conditioning sensor signals based on neural networks

Sensor signal conditioning represents the front-end of a data acquisition system. For this purpose, some specific electronic components are required. In industrial electronics applications the final system response should be independent of variations commonly present in the sensor output (due to fabrication process, tolerances, drifts, etc.). This paper shows a sensor conditioning procedure, based on a small neural network programmed onto a low-cost standard microcontroller. Due to its flexibility, this technique improves the sensor response, providing the same output values even for sensors with different behaviour, and without additional electronic components. Results, execution times and hardware requirements are presented.