Radiometer design analysis based upon measurement uncertainty

This dissertation introduces a method for predicting the performance of a radiometer design based on calculating the measurement uncertainty. The variety in radiometer designs and the demand for improved radiometric measurements justify the need for a more general and comprehensive method to assess system performance. Radiometric resolution, or sensitivity, is a figure of merit that has been commonly used to characterize the performance of a radiometer. However when evaluating the performance of a calibration design for a radiometer, the use of radiometric resolution has limited application. These limitations are overcome by considering instead the measurement uncertainty. A method for calculating measurement uncertainty for a generic radiometer design including its calibration algorithm is presented. The technique entails decomposing a radiometer design into a set of subsystems from which a set of characteristic equations can be derived. The set of equations are used to form an estimator for the radiometer input. The expected value of the mean square error of the estimator is the measurement uncertainty and is the basis for quantitative comparative analysis. The result is a generalized technique by which system calibration architectures and design parameters can be evaluated to optimize instrument performance for given requirements and constraints. Least squares regression is presented as a framework by which a wide variety of calibration schemes can be analyzed. Example applications demonstrate the utility of using measurement uncertainty as a figure of merit. An experiment was conducted to validate some model predictions. Comparisons of model calculations with measurements using different calibration algorithms demonstrate the viability of the technique.