The University of Klagenfurt and Infineon Technologies Austria have in cooperation developed a toolbox to enable easy modelling and calculations for applying uncertain parameters. The tooling is implemented object oriented and uses operator overloading to simplify working with uncertainties in the MATLAB toolset which is well known by many design engineers and system architects.
Sensor systems provide estimates of quantities of interest. These estimates are subject to some degree of uncertainty. Design Engineers of such sensor systems need to develop systems with known low uncertainties and need to ensure a maximum production yield while also considering safety aspects. With this toolbox it is possible to define uncertain variables with mean value and its associated standard-deviation and model mathematical algorithms, where mean-values and standard uncertainties are propagated through the mathematical model using the first order approximations, which is accurate for small deviations such as are expected in a sensor system. Even correlations of parameter variations as well as complex values can be propagated. The knowledge of system output uncertainties even in early architectural phase enables fast decisions and optimizations as well as reduction of expensive design iterations by measurement of the architectural performance after production and characterization.
As use-case example, temperature measurement system was used to show how to include system uncertainties in early design phases as well as in the process of system calibration and optimization
This figure shows the non-optimized and optimized calibrated sensor performance. In this sensor system example it could be shown how this toolbox can help to statistically optimize this temperature sensor system.