Deconvolution Based on Experimentally Determined Apparatus Functions

Deconvolution of experimental measurements in e.g. electron or ion scattering can result in considerable resolution enhancement. It is usually assumed that the apparatus function which we wish to remove from the experimental signal is either known exactly or with a much higher precision than the signal. This assumption is not valid in general. In fact many situations are conceivable where measurement of the apparatus function requires the same effort as measurement of the signal. We have performed a rigorous Bayesian analysis for this general case and present applications to Rutherford backscattering from thin films.