On the implementation of multiple delay line filters with digital signal processing techniques

By using an analog prefilter and a sampled data processor, it is possible to easily implement an entire class of weighting functions. Their main feature is a finite time duration; an additional interesting property is the small number of samples needed to process a pulse. Flat tops within better than 1% accuracy can also be achieved. Possible implementations of these filters are discussed with emphasis on causes of inaccuracy. In particular, these filters come out to be time-variant. Synchronization issues are therefore developed and thoroughly discussed. In a mixed analog-digital implementation, special regard should be given to the A/D converter: the requirements on its characteristics are derived. In particular, a simple way to evaluate its resolution (number of bits) is derived, and the interrelationships between the achievable accuracy, ballistic deficit reduction, and the characteristics of the analog prefilter are discussed. Strategies aimed at achieving and/or at avoiding the synchronization of the input pulses with the digital acquisition are proposed.<
>