Quantitative evaluation of self compensation algorithms applied in intelligent sensors

Self compensation process in intelligent sensors is important in order to fix offsets, gain, linearity and cross-sensitivity errors. Today a lot of methods to do the compensation are available, but the designer has the problem to determine or select which will be the best method due to the lack of information to know if a compensation method will work well with a particular sensor. In this paper a methodology to make a quantitative evaluation of any compensation method to be used in an intelligent sensor is presented. The designer just needs to know the maximum value of nonlinearity of his sensor. The methodology was simulated using four compensation methods: piecewise, polinomial progressive, improved polinomial progressive and artificial neural networks. To validate these methods sensors with the worst nonlinearity were used, like thermistor and one distance sensor. The results are summarized in tables in order to facilitate their use. The objective of the proposal methodology is to save designing time and calibration costs because the designer could easily chose the algorithm that requires minimum readjustment points.