Nichtlineare Approximationsmethoden zur Reduzierung nichtidealer Sensoreigenschaften in integrierten CMOS-Sensorsystemen

For modern instrumentation systems, sensors obtain a special significance, since they are necessary interfaces to technical processes. These sensors are responsible for conversion of nonelectric measurands into electrical signals. For this conversion sensors use different physical measuring effects, which are often nonlinear. Furthermore these effects are additionally sensitive to several disturbance quantities. Moreover they depend on drift effects caused by aging as well as statistical material and process variations. Hence a proper temperature compensation of these unwanted effects and a calibration of nominal sensing parameters is essential for precise and high selective measurements. For this the sensor must be applied to defined nonelectric reference quantities and must be adjusted using suitable trimming elements, so that the static transfer function stays inside a given tolerance under specified influence or disturbance quantity. The presented thesis employs conventional as wel l as novel approximation procedures for cost-efficient reduction of nonideal sensor properties and finally compares their advantages and disadvantages. To these procedures belong a simple look-up table concept which use the interpolating properties of oversampled A/D-converter. With this novel concept the spectrum of a piecewise linear, square, and cubic polynomial interpolation can be realised. The novel concept is also used for efficient temperature compensation of a smart piezoresistive pressure transducer. Thus the heavy temperature dependence of pressure transducer can be reduced below 100ppm/K. The high flexibility of this concept can also be used for many other problems in sensor development.