Calorimeter Signal Noise Analysis and Sample Power Prediction Improvement

This paper presents some results from an analysis directed towards improving the sample power end-point prediction algorithms (predictors) used with isothermal calorimeters. Simulated and real measurements are given which show how the accuracy of the sample power end-point predictions can be improved. Also the measurement time needed by the predictor to estimate the sample power is reduced, leading to shorter assay times. The improvements in the end-point prediction accuracy are a result of using powerful digital filters. The filters were designed after an analysis of the nature of the noise present on the calorimeter measurement chamber power data. The use of end-point prediction algorithms based on fitting the calorimeter response to a single decaying exponential equation are well established. By extending the fitting function to a double exponential it is possible to estimate the sample power earlier in the measurement cycle, thus reducing the measurement time. Real data is presented which shows the potential of such fitting algorithms by yielding a 50% decrease in measurement time compared to the single exponential end-point predictor.