The quality of a measuring system depends more and more on the first stages of the measuring chain i.e. the sensor and the primary signal conditioning circuit. The use of a high resolution A/D converter at the very beginning of that chain is generally correct. In practice however, sensors are mainly analogue and they need some analogue signal conditioning circuits. The paper deals with the analogue signal processing in the circuits with a single passive sensor. Uncertainty, error elimination and a nonlinearity of some essential structures have been analysed. The general conclusions may be formulated as follows: Extremely simple methods of analogue signal processing lead to quite good results but some limitations have to be respected. The passive sensor parameter which determines almost all limitations is the ratio ÄR_X / R₀.

Practical usefulness of the result of spectrometric data correction depends considerably on how precise its uncertainty is assessed. If the algorithm used for correction is linear with respect to the data and all intermediate results of computation, and if it does not use the data or intermediate results of computation more than once, then the linear propagation of uncertainties represented by intervals, executed step by step for each operation, yields a satisfactory estimate of the measurement uncertainty. Otherwise, a tendency to overestimate the uncertainty appears. An effective interval-based method for dealing with this problem in weakly non-linear algorithms has been recently proposed by the authors. In this paper, it is implemented in a new algorithm for spectrometric data correction. Computational efficiency of this algorithm is demonstrated using synthetic spectrometric data.

The main problems related with taking into account the a priori knowledge about the measured signal have been outlined. Then two examples have been presented, showing how the knowledge about one signal parameter narrows down the uncertainty about the other one. Firstly, bounding curves for the meansquare value when the mean value is known have been examined and some practical conclusions concerning single and multiple signal rectification have been presented. Then the mean value uncertainty for a given signal bandwidth has been studied.

The contribution deals with evaluation of uncertainties in the process of calibration of multicomponent sensors with special aim to force matrix sensor arrays. Matrix sensors are designed for simultaneous measurement of acting forces. It is used in applications in which sensing of varying forces acting in an area is needed (e.g. in robotics, telemanipulating devices, medical applications). Model for calibration of matrix sensor arrays and the methodology for uncertainties and covariances calculation is submitted. The base is regress polynomial model of calibration. The result is estimation of model parameters and covariance matrix of those estimators. Considered are all uncertainties and covariances, resulting from measuring standard device, transmission of values from measuring standard to sensor, repeating of values by calibrating sensor. The most important characteristics of current stateof- art in this field is absence of unique procedures used for objective matrix sensor arrays evaluation and testing.

In the article there has been presented the methodology of working out results of mutually coupled measurements of shape deviations and axis location of large engine elements carried out by non-referential methods. There has also been presented a way of estimating measurement errors of shape deviations and axis location deviations of cylindrical outer and inner area complexes.