Stochastic digital filtering of nuclear radiation signals : a new approach

We present a new approach for the design of optimum filters with good low frequency rejection suitable for high resolution nuclear spectrometry. The method, based on digital filtering techniques, permits to well approximate the theoretical signal-to-noise ratio limit while keeping a good rejection to slow disturbing waveforms. The proposed filter weight function uses the `radiation-free' time intervals closely spaced around the signal pulse for optimally estimating the baseline level and subtracts the obtained value from the pulse height estimate. We show that the resolution loss due to the `optimum baseline restoration' is of the order of a few percent even in case of high repetition rates, with a considerable improvement compared to an algorithm of simple average of the baseline samples. Because the occurrence of the radiation free time intervals is generally subject to a Poissonian statistic, the filter must be adaptive, which implies that the overall weight function is stochastic. The low-frequency rejection for this stochastic filter has been determined as a statistic property, resulting from the average of the used filter frequency responses