IEEE 1451 Correction Engine to Temperature-compensation of Magnetoresistive Transducers
Helena G. Ramos, Pedro Silva Girão, Octavian Adrian Postolache, J. M. Dias Pereira · 2006
The paper presents the comparison between polynomial approximation and artificial neural networks (ANNs) to compensate temperature dependence of a magnetic field transducer. The sensing elements are a magnetoresistance whose value can vary almost 20% in the experimental operating temperature range (20degC-100degC) and a two terminal integrated temperature sensor. The first technique to correct the temperature drift in the magnetoresistance is fully compliant with IEEE 1451.2 correction engine. It uses a segmented multinomial (multivariate polynomial) function and the coefficients and offset values stored in TEDS are determined using a least-mean-square error method. The application of an artificial neural network, well adapted to conveniently modeling strongly nonlinear transducer characteristics, is the second technique to be used and leads to an improvement of magnetic transducer's accuracy from 20% to 2%. An approach to a "correction engine" covering this method is proposed